Electronic Commerce Research and Applications
Volume 6 • Issue 6
- ISSN: 1567-4223
Editor-In-Chief: Christopher Yang
- 5 Year impact factor: 6.9
- Impact factor: 5.9
- Journal metrics
Electronic Commerce Research and Applications aims to create and disseminate enduring knowledge for the fast-changing e-commerce environment. A major dilemma in e-commerce resea… Read more
Subscription options
Institutional subscription on sciencedirect.
Electronic Commerce Research and Applications aims to create and disseminate enduring knowledge for the fast-changing e-commerce environment. A major dilemma in e-commerce research is how to achieve a balance between the currency and the life span of knowledge.
Electronic Commerce Research and Applications will contribute to the establishment of a research community to create the knowledge, technology, theory, and applications for the development of electronic commerce . This is targeted at the intersection of technological potential and business aims.
E-commerce is a multi-disciplinary area, which should be developed in co-operation with existing fields such as Information Systems and Technology; Computing and Informatics; Marketing, Finance and Supply Chain Management; Business Strategy and Management; Artificial Intelligence/Machine Learning; Data Science and Business Data Analytics; Public Policy; and Legal Studies. We will solicit papers on current technologies from these areas, as well as publish papers on completely new topics. We also seek proposals for special issues on new topics in e-commerce that will create new directions for research.
Electronic Commerce Research and Applications is inviting submission of articles, including but not limited to the following topics:Agent-based commerce; electronic auctions; e-business models; B2C and B2B EC; consumer behavior; customer relationship management and data mining; recommender systems; Internet search engines and Web mining; big data analytics; social media and commerce analytics; responsible and trustworthy artificial intelligence; pricing and marketing; digital economy and digital transformation; e-government, public policy and digital divide issues; electronic payment systems; sharing economy; (IT and e-services; exchanges and electronic marketplaces;) e-commerce in supply chain and inventory management; legal issues in e-commerce; (industry studies and case analysis;) economic and management science modeling; organizational and theory-building research; empirical studies of e-commerce problems; behavioral studies of e-commerce issues; protocols, technology and process standards for e-commerce; (transformation of industries;) security and trust; credit card and smart card applications; mobile-commerce and ubiquitous computing; inter-organizational systems in e-commerce; emerging technologies and technological innovation.
Electronic Commerce Research and Applications
Volume 66, Issue C
- Elsevier Science Publishers B. V.
- PO Box 211 1000 AE Amsterdam
- Netherlands
Subject Areas
Most frequent affiliations, most cited authors, latest issue.
- Volume 66, Issue C Jul 2024 ISSN: 1567-4223 View Table of Contents
Editorial Board
How perceived justice leads to stickiness to short-term rental platforms: unveiling the effect of relationship commitment and trust.
Renmin University of China, Beijing, China
University of International Business and Economics, Beijing, China
Information transmit strategy of e-commerce platform with financially constrained supplier
Research Center for Innovation and Development of Equipment Manufacturing Industry, Taiyuan University of Science and Technology, Taiyuan 030024, China
College of Economy and Management, Taiyuan University of Science and Technology, Taiyuan 030024, China
Research on the joint event extraction method orientates food live e-commerce
Beijing Key Laboratory of Big Data Technology for Food Safety, Beijing Technology and Business University, Beijing 100048, China
Sub-institute of Standardization Theory and Strategy, China National Institute of Standardization(CNIS), Beijing 100088, China
The Law Promotion Center of the Ministry of Justice of the People’s Republic of China, Beijing 100020, China
Research Center of Information Technology, Beijing Academy of Agriculture and Forestry Sciences, Beijing 100097, China
Beijing PAIDE Science and Technology Development Co. Ltd., Beijing 100097, China
Improved negative sampling method in collaborative filtering recommendation based on Generative adversarial network
College of Artificial Intelligence, Nanjing University of Posts and Telecommunications, Nanjing 210023, China
College of Information Engineering, Nanjing University of Finance & Economics, Nanjing 210023, China
Recent Award Winners
Most popular, export citations.
- Please download or close your previous search result export first before starting a new bulk export. Preview is not available. By clicking download, a status dialog will open to start the export process. The process may take a few minutes but once it finishes a file will be downloadable from your browser. You may continue to browse the DL while the export process is in progress. Download
- Download citation
- Copy citation
We are preparing your search results for download ...
We will inform you here when the file is ready.
Your file of search results citations is now ready.
Your search export query has expired. Please try again.
Economics and Electronic Commerce
Google is a global leader in electronic commerce. Not surprisingly, considerable attention is devoted to research in this area. Topics include 1) auction design, 2) advertising effectiveness, 3) statistical methods, 4) forecasting and prediction, 5) survey research, 6) policy analysis and a host of other topics. This research involves interdisciplinary collaboration among computer scientists, economists, statisticians, and analytic marketing researchers both at Google and academic institutions around the world.
A major challenge is solving these problems at very large scales. For example, the advertising market has billions of transactions daily, spread across millions of advertisers, presenting a unique opportunity to test and refine economic principles as applied to a very large number of interacting, self-interested parties with a myriad of objectives. At Google, research is an opportunity to translate direction into practice, influencing how production systems are designed and used.
Recent Publications
Some of our teams.
Algorithms & optimization
Graph mining
Impact-Driven Research, Innovation and Moonshots
Market algorithms
We're always looking for more talented, passionate people.
An official website of the United States government
The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.
The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.
- Publications
- Account settings
Preview improvements coming to the PMC website in October 2024. Learn More or Try it out now .
- Advanced Search
- Journal List
- Springer Nature - PMC COVID-19 Collection
20 years of Electronic Commerce Research
Satish kumar.
1 Department of Management Studies, Malaviya National Institute of Technology, Jaipur, Rajasthan 302017 India
Weng Marc Lim
2 Swinburne Business School, Swinburne University of Technology, John Street, Hawthorn, Victoria 3122 Australia
3 School of Business, Swinburne University of Technology, Jalan Simpang Tiga, 93350 Kuching, Sarawak Malaysia
Nitesh Pandey
J. christopher westland.
4 Department of Information and Decision Sciences, University of Illinois – Chicago, 601 S. Morgan Street, Chicago, Illinois 60607-7124 USA
2021 marks the 20th anniversary of the founding of Electronic Commerce Research ( ECR ). The journal has changed substantially over its life, reflecting the wider changes in the tools and commercial focus of electronic commerce. ECR ’s early focus was telecommunications and electronic commerce. After reorganization and new editorship in 2014, that focus expanded to embrace emerging tools, business models, and applications in electronic commerce, with an emphasis on the innovations and the vibrant growth of electronic commerce in Asia. Over this time, ECR ’s impact and volume of publications have grown rapidly, and ECR is considered one of the premier journals in its discipline. This invited research summarizes the evolution of ECR ’s research focus over its history.
Introduction
The year 2021 marks the 20th anniversary of the founding of Electronic Commerce Research ( ECR ). The journal has changed substantially over its life, reflecting the wider changes in the tools and commercial focus of electronic commerce. ECR ’s early focus was on telecommunications and electronic commerce. After reorganization and new editorship in 2014, that focus expanded to embrace emerging tools, business models, and applications in electronic commerce, with an emphasis on emerging technologies and the vibrant growth of electronic commerce in Asia. Over these years, ECR has steadily improved its stature and impact, as evidenced through various quantitative (e.g., citations, impact factors) and qualitative (e.g., peer-informed journal ranks) measures. According to Clarivate Analytics, ECR ’s impact factor in 2019 was 2.507, 1 which means that articles published in ECR between 2017 and 2018 received an average of 2.507 citations from journals indexed in Web of Science in 2019. The five-year impact factor of ECR was 2.643, 1 which indicates that articles published in ECR between 2014 and 2018 received an average of 2.643 citations from Web of Science-indexed journals in 2019. According to Scopus, ECR ’s CiteScore was 4.3, 2 which implies that articles published in ECR between 2016 and 2019 received an average of 4.3 citations from journals indexed in Scopus in 2019. The source normalized impact per paper (SNIP) of ECR was 1.962, which suggests that the average citations received by articles in the journal is 1.962 times the average citations received by articles in the same subject area of Scopus-indexed journals in 2019. Apart from these quantitative measures, ECR has also been rated highly by peers in the field, as seen through journal quality lists. For example, ECR has been consistently ranked as an “A” journal by the Excellence in Research for Australia (ERA 2010) and the Australian Business Deans Council (ABDC 2013, 2016, 2019) journal ranking lists.
This research presents a 20-year retrospective bibliometric analysis of the evolution of context and focus of ECR ’s articles [ 1 – 5 ]. To curate a rich bibliometric overview of ECR ’s scientific achievements, this study explores seven research questions (RQ) which are commonly asked by both authors and our Editorial Board members:
- RQ1. What is the trend of publication and citation in ECR ?
- RQ2. Who are the most prolific contributors (authors, institutions, and countries) in ECR ?
- RQ3. What are the most influential publications in ECR ?
- RQ4. Where have ECR publications been cited the most?
- RQ5. What is the trend of collaboration in ECR ?
- RQ6. Who are the most important constituents of the collaboration network in ECR ?
- RQ7. What are the major research themes in ECR ?
A bibliometric analysis can offer a broad, systematic overview of the literature to delineate the evolution of electronic commerce technologies, and point the direction to trending topics and methodologies [ 5 – 14 ]. Our research is organized as follows. Section 2 outlines our bibliometric methodology. Section 3 goes on to performance analysis to uncover contributor and journal performance trends (RQ1–RQ4), the co-authorship analysis performed to unpack collaboration and constituent characteristics (RQ5–RQ6), and the bibliometric coupling and keyword analyses used to reveal the major themes and trends within the ECR corpus (RQ7). Section 4 applies graph theoretic analysis. Section 5 applies cluster analysis. Section 6 applies thematic analysis. Finally, we conclude the study with key takeaways from this retrospective.
Methodology
Bibliometric methodologies apply graph theoretic and statistical tools for analysis of bibliographic data [ 15 ] and include performance analysis and science mapping [ 16 ]. To answer research question 1 to research question 4, this study uses performance analysis to measure the output of authors’ productivity and impact, with productivity measured using publications per year, and impact measured using citations per year. We begin by measuring the productivity and impact of ECR , and then the productivity and impact of authors, institutions, and countries using both publications and citations per year metrics on top of ancillary measures such as citations per publication and h -index. Finally, we measure the impact of ECR articles using citations and shed light on prominent publication outlets citing ECR articles.
To answer research question 5 to research question 7, this study uses co-authorship, bibliographic coupling, and keyword analyses. We begin by conducting a co-authorship analysis, which is a network-based analysis that scrutinizes the relationships among journal contributors [ 17 ]. Next, we perform bibliographic coupling to obtain the major themes within the ECR corpus. The assumption of bibliographic coupling connotes that two documents would be similar in content if they share similar references [ 18 , 19 ]. Using article references, a network was created, wherein shared references were assigned with edge weights and documents were denoted with nodes. The documents were divided into thematic clusters using the Newman and Girvan [ 20 ] algorithm. Finally, we track the development of themes throughout different time periods using a temporal keyword analysis. The assumption of this analysis suggest that keywords are representative of the author’s intent [ 21 ] and thus important for understanding the prominence of themes pursued by authors across different time periods. Indeed, we found that these bibliometric methods complement each other relatively well, as bibliographic coupling was useful to locate general themes while keywords were useful to understand specific topics.
To acquire bibliographic data of ECR articles for the bibliometric analyses mentioned above, this study uses the Scopus database, which is one of the largest academic database that is almost 60% larger than the Web of Science [ 21 ]. Past research has also indicated that the citations presented within the Scopus database correlate more with expert judgement as compared to Google Scholar and Web of Science [ 22 ]. We begin by conducting a source search for “ Electronic Commerce Research ,” which resulted in 927 articles, and after filtering out non- ECR articles, we obtain a list of 516 ECR articles (see Fig. 1 ). However, ECR only gained Scopus indexation in 2005, and thus, only 443 ECR articles (2005–2020) contained full bibliometric data, whereas the remaining 73 ECR articles (2001–2004) contained only partial bibliometric data (e.g., no affiliation, abstract, and keyword entry). All 516 ECR articles were fetched and included in the performance analysis as partial bibliometric data was sufficient, but only 443 ECR articles were included in science mapping (e.g., co-authorship, bibliographic coupling, and keyword analyses using VOSviewer [ 23 ] and Gephi [ 24 ]) as full bibliometric data was required. This collection of articles met the minimum sample size of 200 articles for bibliometric analysis recommended by Rogers, Szomszor, and Adams [ 25 ].
Research design. Note Bibliometric analysis was conducted for only 443 (primary) documents as 73 (secondary) documents lack full data (affiliation, abstract and keywords)
Performance analysis: productivity and impact
The publication and citation trends of ECR between 2001 and 2020 are presented in Fig. 2 (RQ1). In terms of publication, the number of articles published in ECR has grown from 20 articles per year in 2001 to 81 articles per year in 2020, with an average annual growth rate of 7.64%. In terms of citations, the number of citations that ECR articles received has grown from three citations in 2001 to 1219 citations in 2020, with an average annual growth rate of 37.19%. These statistics suggest that ECR ’s publications and citations have seen exponential growth since its inception, and that the journal’s citations have grown at a much faster rate than its publication, which is very positive.
Annual publication and citation structure of ECR
The most prolific authors in ECR between 2001 and 2020 are presented in Table Table1 1 (RQ2). The most prolific author is Jian Mou, who has published six articles in ECR , which have garnered a total of 95 citations. This is followed by Yan-Ping Liu and Liyi Zhang, who have published three articles each in ECR , which have received a total of 46 and 42 citations, respectively. Among the top 20 contributors, the author with the highest citation average per publication is Katina Michael (TC/TP and TC/TCP = 59 citations), who is followed closely by Yue Guo (TC/TP and TC/TCP = 51 citations); they are the only two authors who have an average citation greater than 50 for their ECR articles.
Most prolific authors for ECR between 2001 and 2020
Author | TP | TCP | TC | TC/TP | TC/TCP | |
---|---|---|---|---|---|---|
Mou J | 6 | 5 | 95 | 15.83 | 19.00 | 4 |
Liu Y.-P | 3 | 3 | 46 | 15.33 | 15.33 | 3 |
Zhang L | 3 | 2 | 42 | 14.00 | 21.00 | 2 |
Lin Z | 3 | 3 | 40 | 13.33 | 13.33 | 3 |
Westland J.C | 3 | 3 | 21 | 7.00 | 7.00 | 3 |
Luo X | 3 | 1 | 19 | 6.33 | 19.00 | 1 |
Yan B | 3 | 2 | 6 | 2.00 | 3.00 | 1 |
Sun J | 3 | 0 | 0 | 0.00 | 0.00 | 0 |
Michael K | 2 | 2 | 118 | 59.00 | 59.00 | 2 |
Guo Y | 2 | 2 | 102 | 51.00 | 51.00 | 2 |
Choo K.-K.R | 2 | 2 | 78 | 39.00 | 39.00 | 2 |
Khedmatgozar H.R | 2 | 2 | 74 | 37.00 | 37.00 | 2 |
Wei J | 2 | 2 | 73 | 36.50 | 36.50 | 2 |
Teng C.-I | 2 | 2 | 71 | 35.50 | 35.50 | 2 |
Paraschiv C | 2 | 2 | 67 | 33.50 | 33.50 | 2 |
Chen M.-Y | 2 | 2 | 63 | 31.50 | 31.50 | 2 |
Cohen J | 2 | 2 | 61 | 30.50 | 30.50 | 2 |
Maes P | 2 | 2 | 56 | 28.00 | 28.00 | 2 |
Tsao W.-C | 2 | 2 | 55 | 27.50 | 27.50 | 2 |
Lee H.S | 2 | 2 | 50 | 25.00 | 25.00 | 2 |
TP = total publication(s). TCP = total cited publication(s). TC = total citation(s). TC/TP = cites per publication. TC/TCP = cites per cited publication. h = h -index
Institutions
The most prolific institutions for ECR between 2001 and 2020 are presented in Table Table2 2 (RQ2). IBM, with 14 articles and 371 citations, emerges as the highest contributing institution to ECR . It is surprising yet encouraging to see a high number of contributions coming from practice, which reflects the ECR ’s receptiveness to publish industry-relevant research. Nonetheless, it is worth mentioning that this contribution is derived from the collective effort of IBM’s research labs around the world (e.g., Delhi, Haifa, and New York)—a unique advantage that most higher education institutions do not enjoy unless they have full-fledged research-active international branch campuses around the world. The second and third most contributing institutions are Nanjing University and Xi’an Jiaotong University, with 11 and 10 articles that have been cited 116 and 29 times, respectively. This is yet another interesting observation, as the contributions by Chinese institutions suggest that ECR is a truly international journal despite its origins and operations stemming in the United States. Finally, the University of California (TC/TP and TC/TCP = 34.86 citations) emerges as the institution that averages the most citations per publication, followed by IBM (TC/TP and TC/TCP = 26.50 citations) and Texas Tech University (TC/TP and TC/TCP = 26.20 citations).
Most prolific institutions for ECR between 2001 and 2020
Institution | TP | TCP | TC | TC/TP | TC/TCP | |
---|---|---|---|---|---|---|
IBM | 14 | 14 | 371 | 26.50 | 26.50 | 9 |
Nanjing University | 11 | 7 | 116 | 10.55 | 16.57 | 4 |
Xi'an Jiaotong University | 10 | 7 | 29 | 2.90 | 4.14 | 3 |
Zhejiang University | 9 | 8 | 85 | 9.44 | 10.63 | 5 |
Xidian University | 8 | 6 | 56 | 7.00 | 9.33 | 4 |
Shanghai University | 8 | 2 | 16 | 2.00 | 8.00 | 2 |
University of California | 7 | 7 | 244 | 34.86 | 34.86 | 7 |
University of Texas | 7 | 6 | 94 | 13.43 | 15.67 | 4 |
Wuhan University | 7 | 5 | 68 | 9.71 | 13.60 | 4 |
Hefei University of Technology | 7 | 2 | 8 | 1.14 | 4.00 | 2 |
Queensland University of Technology | 6 | 6 | 138 | 23.00 | 23.00 | 4 |
Tsinghua University | 6 | 4 | 57 | 9.50 | 14.25 | 4 |
South China University of Technology | 6 | 4 | 44 | 7.33 | 11.00 | 3 |
Soochow University | 6 | 6 | 40 | 6.67 | 6.67 | 3 |
University of Illinois | 6 | 6 | 31 | 5.17 | 5.17 | 4 |
Texas Tech University | 5 | 5 | 131 | 26.20 | 26.20 | 5 |
University of Wisconsin | 5 | 5 | 99 | 19.80 | 19.80 | 4 |
Victoria University of Wellington | 5 | 5 | 90 | 18.00 | 18.00 | 4 |
City University of Hong Kong | 5 | 5 | 54 | 10.80 | 10.80 | 4 |
University of Alabama | 5 | 5 | 29 | 5.80 | 5.80 | 4 |
The most prolific countries in ECR between 2001 and 2020 are presented in Table Table3 3 (RQ2). China emerges as the most prolific contributor, with 152 articles and 1066 citations. This is followed by the United States, which has contributed 143 articles and 2813 citations. No country other than China and the United States has contributed more than 50 articles to ECR . Nevertheless, it is important to note that ECR also receives contributions from many countries around the world, as the remaining ± 50% of contributions in the top 20 list comes from 18 different countries across Asia, Europe, and Oceania.
Most prolific countries for ECR between 2001 and 2020
Country | TP | TCP | TC | TC/TP | TC/TCP | |
---|---|---|---|---|---|---|
China | 152 | 108 | 1066 | 7.01 | 9.87 | 15 |
United States | 143 | 133 | 2813 | 19.67 | 21.15 | 23 |
Taiwan | 35 | 34 | 535 | 15.29 | 15.74 | 12 |
South Korea | 34 | 26 | 451 | 13.26 | 17.35 | 7 |
Australia | 30 | 28 | 547 | 18.23 | 19.54 | 13 |
Germany | 21 | 20 | 565 | 26.90 | 28.25 | 9 |
United Kingdom | 21 | 18 | 413 | 19.67 | 22.94 | 11 |
India | 21 | 15 | 163 | 7.76 | 10.87 | 6 |
Spain | 20 | 18 | 321 | 16.05 | 17.83 | 12 |
Greece | 17 | 17 | 512 | 30.12 | 30.12 | 11 |
Hong Kong | 16 | 13 | 142 | 8.88 | 10.92 | 8 |
Canada | 15 | 12 | 506 | 33.73 | 42.17 | 10 |
France | 13 | 11 | 218 | 16.77 | 19.82 | 9 |
Italy | 13 | 11 | 186 | 14.31 | 16.91 | 8 |
Switzerland | 11 | 11 | 250 | 22.73 | 22.73 | 9 |
Iran | 11 | 11 | 222 | 20.18 | 20.18 | 7 |
New Zealand | 10 | 10 | 167 | 16.70 | 16.70 | 8 |
Japan | 8 | 7 | 87 | 10.88 | 12.43 | 5 |
Singapore | 7 | 7 | 215 | 30.71 | 30.71 | 5 |
Sweden | 6 | 6 | 79 | 13.17 | 13.17 | 4 |
The most cited articles in ECR between 2001 and 2020 are presented in Table Table4 4 (RQ3). The most cited article published in ECR during this period is Füller et al.’s [ 26 ] article on the role of virtual communities in new product development (TC = 270). This is followed by Sotiriadis and van Zyl’s [ 27 ] article on electronic word of mouth and its effects on the tourism industry (TC = 188), Nonnecke et al.’s [ 28 ] article on the phenomena of ‘lurking’ in online communities (TC = 185), Lehdonvirta’s [ 29 ] article on the factors that drive virtual product purchases (TC = 170), and Bae and Lee’s [ 30 ] article on the effect of gender on consumer perception of online reviews (TC = 125). The diversity of topics in the most cited articles indicate that electronic commerce is indeed a multi-faceted subject, which we will explore in detail in the later sections.
Most cited articles published in ECR between 2001 and 2020
Author(s) | Title | Year | TC | C/Y |
---|---|---|---|---|
Füller J., Bartl M., Ernst H., Mühlbacher, H | Community based innovation: how to integrate members of virtual communities into new product development | 2006 | 270 | 18.00 |
Sotiriadis M.D., van Zyl C | Electronic word-of-mouth and online reviews in tourism services: the use of Twitter by tourists | 2013 | 188 | 23.50 |
Nonnecke B., Andrews D., Preece J | Non-public and public online community participation: needs, attitudes and behavior | 2006 | 185 | 12.33 |
Lehdonvirta V | Virtual item sales as a revenue model: identifying attributes that drive purchase decisions | 2009 | 170 | 14.17 |
Bae S., Lee T | Gender differences in consumers’ perception of online consumer reviews | 2011 | 125 | 12.50 |
Kim J.B | An empirical study on consumer first purchase intention in online shopping: integrating initial trust and TAM | 2012 | 124 | 13.78 |
Zarmpou T., Saprikis V., Markos A., Vlachopoulou M | Modeling users’ acceptance of mobile services | 2012 | 121 | 13.44 |
Sila I | Factors affecting the adoption of B2B e-commerce technologies | 2013 | 118 | 14.75 |
Malaga R.A | Web-based reputation management systems: problems and suggested solutions | 2001 | 118 | 5.90 |
Gregg D.G., Walczak S | The relationship between website quality, trust and price premiums at online auctions | 2010 | 104 | 9.45 |
Huang T.-L., Liao S | A model of acceptance of augmented-reality interactive technology: the moderating role of cognitive innovativeness | 2015 | 102 | 17.00 |
Flanagin A.J., Metzger M.J., Pure R., Markov A., Hartsell E | Mitigating risk in ecommerce transactions: perceptions of information credibility and the role of user-generated ratings in product quality and purchase intention | 2014 | 101 | 14.43 |
Lee P.M | Behavioral model of online purchasers in e-commerce environment | 2002 | 101 | 5.32 |
Guo Y., Barnes S | Virtual item purchase behavior in virtual worlds: an exploratory investigation | 2009 | 97 | 8.08 |
Pourshahid A., Amyot D., Peyton L., Ghanavati S., Chen P., Weiss M., Forster A.J | Business process management with the user requirements notation | 2009 | 96 | 8.00 |
Hsieh J.-K., Hsieh Y.-C., Tang Y.-C | Exploring the disseminating behaviors of eWOM marketing: persuasion in online video | 2012 | 80 | 8.89 |
Xu F., Michael K., Chen X | Factors affecting privacy disclosure on social network sites: an integrated model | 2013 | 79 | 9.88 |
Patton M.A., Josang A | Technologies for trust in electronic commerce | 2004 | 79 | 4.65 |
Taylor D.G., Davis D.F., Jillapalli R | Privacy concern and online personalization: the moderating effects of information control and compensation | 2009 | 76 | 6.33 |
Jeffrey S.A., Hodge R | Factors influencing impulse buying during an online purchase | 2007 | 74 | 5.29 |
TC = total citation(s). C/Y = cites per year
Publication outlets
The publication outlets that have cited ECR articles the most between 2001 and 2020 are presented in Table Table5 5 (RQ4). The list includes many prestigious journals such as International Journal of Information Management (ABDC = A*, IF = 8.210), Information and Management (ABDC = A*, IF = 5.155), and Decision Support Systems (ABDC = A*, IF = 4.721), among others. The presence of such reputed journals reflects ECR ’s own reputation of high standing among its peers. Apart from ECR , the publication outlets that have highly cited ECR include Lecture Notes in Computer Science including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics (TC = 218), Computers in Human Behavior (TC = 95), and ACM International Conference Proceeding Series (TC = 88), which reflect the diversity in publication outlets that ECR is making an impact (e.g., book, conference, journal).
Publications citing ECR the most between 2001 and 2020
tTitle | TC | ABDC rank | IF | CiteScore | SNIP |
---|---|---|---|---|---|
Electronic Commerce Research | 267 | A | 2.507 | 4.3 | 1.962 |
Lecture Notes in Computer Science including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics | 218 | NA | NA | NA | NA |
Computers in Human Behavior | 95 | A | 5.003 | 12.1 | 3.079 |
ACM International Conference Proceeding Series | 88 | NA | NA | 0.8 | 0.333 |
Journal of Retailing and Consumer Services | 69 | A | 4.219 | 7.4 | 2.166 |
Sustainability | 67 | NA | 2.567 | NA | NA |
Electronic Commerce Research and Applications | 61 | C | 3.824 | 6.9 | 1.787 |
Advances in Intelligent Systems and Computing | 60 | NA | NA | 0.9 | 0.429 |
Lecture Notes in Business Information Processing | 51 | NA | NA | 1.3 | 0.573 |
Internet Research | 47 | A | 4.708 | 7.9 | 2.213 |
Journal of Business Research | 44 | A | 4.874 | 8.9 | 2.76 |
International Journal of Information Management | 43 | A* | 8.21 | 14.1 | 3.773 |
Decision Support Systems | 43 | A* | 4.721 | 9.5 | 2.718 |
IEEE Access | 41 | NA | 3.745 | 3.9 | 1.734 |
Information and Management | 38 | A* | 5.155 | 11 | 3.002 |
Communications in Computer and Information Science | 37 | NA | NA | 0.7 | 0.403 |
Ceur Workshop Proceedings | 31 | NA | NA | 0.6 | 0.293 |
Journal of Electronic Commerce Research | 30 | B | 1.875 | 4 | 0.963 |
Industrial Management and Data Systems | 30 | A | 3.329 | 7.9 | 2.502 |
Telematics and Informatics | 29 | C | 4.139 | 9.7 | 2.566 |
Journal of Internet Commerce | 29 | B | NA | 3.7 | 1.203 |
Expert Systems with Applications | 29 | C | 5.452 | 11 | 3.139 |
TC = total citation(s). ABDC rank = Australian Business Deans Council rank. IF = 2019 impact factor by Clarivate Analytics. SNIP = 2019 source normalized impact per paper by Scopus. NA = not available
Co-authorship analysis: scientific network
Co-authorship.
The co-authorships in ECR between 2005 and 2020 are presented in Table Table6 6 (RQ5). On the one hand, the co-authorship analysis shows that the share of articles written by a single author has gone down over the years from 10.94% (2005–2008) to 8.61% (2017–2020). The small and decreasing share of single-authored articles do not come as a surprise given the importance and proliferation of collaboration to address increasing thematic and methodological complexity in research [ 31 ]. On the other hand, the co-authorship analysis shows that multi-authored articles have increased their share in ECR , especially articles with three authors or more. In particular, the share of articles with three and five or more authors have increased from 31.25% and 4.69% between 2005 and 2008 to 34.45% and 14.35% between 2017 and 2020, respectively. These statistics suggests that collaboration is growing in prominence, which is consistent with recent observations reported by other premier journals in business [ 32 – 34 ], and that ECR is a good home for collaborative research.
Authors per paper per period
Number of authors | 2005–2008 (%) | 2009–2012 (%) | 2013–2016 (%) | 2017–2020 (%) | Total (%) |
---|---|---|---|---|---|
1 | 10.94 | 17.81 | 12.24 | 8.61 | 11.26 |
2 | 32.81 | 35.62 | 27.55 | 23.44 | 27.70 |
3 | 31.25 | 32.88 | 34.69 | 34.45 | 33.78 |
4 | 20.31 | 5.48 | 14.29 | 19.14 | 15.99 |
≥ 5 | 4.69 | 8.22 | 11.22 | 14.35 | 11.26 |
Network centrality
The most important authors, institutions, and countries across different measures of centrality are presented in Table Table7 7 (RQ6). In this study, we employ four measures of centrality: degree of centrality, betweenness centrality, closeness centrality, and eigen centrality.
Centrality measures for authors, institutions, and countries
Rank | Degree of centrality | Betweenness centrality | Closeness centrality | Eigen centrality |
---|---|---|---|---|
Author | ||||
1 | Mou J | Mou J | Luo X | Wang J.-X |
2 | Sun J | Sun J | Zhang L | Chan F.T.S |
3 | Lin Z | Ding Z | Yan B | Amyot D |
4 | Ding Z | Luo X | Wang Q | Chen P |
5 | Luo X | Zhang L | Chen K | Yang L |
6 | Wang Q | Yan B | Westland J.C | Wang T |
7 | Zhang L | Wang Q | Li X | Choo K.-K.R |
8 | Yan B | Chen K | Zhang N | Weiss M |
9 | Chen K | Zheng H | Fan W | Chen J.-K |
10 | Zheng H | Westland J.C | Kim J | Shi Y |
Institution | ||||
1 | Renmin University | Renmin University | University of Ottawa | University of Ottawa |
2 | University of Ottawa | Xidian University | University of Electronic Science and Technology of China | IBM |
3 | Xidian University | City University of Hong Kong | Yonsei University | Renmin University |
4 | University of Alabama | Chinese Academy of Sciences | Nanjing University | Zhejiang University |
5 | Zhejiang University | University of Ottawa | Beihang University | Hefei University of Technology |
6 | Hefei University of Technology | Beijing Institute Of Technology | Seoul National University | City University of Hong Kong |
7 | University of Texas | University of Illinois | National Cheng Kung University | University of Electronic Science and Technology of China |
8 | Tsinghua University | Wuhan University | National Taichung University | Yonsei University |
9 | IBM | Soochow University | University of the Basque Country | Xi'an Jiaotong University |
10 | City University of Hong Kong | Kookmin University | University of Nottingham | Wuhan University |
Country | ||||
1 | United States | China | United States | United States |
2 | China | United States | Spain | Venezuela |
3 | Spain | Spain | Norway | United Kingdom |
4 | United Kingdom | Greece | Iran | Slovenia |
5 | Australia | Germany | Belgium | South Korea |
6 | Greece | Italy | Czech Republic | Norway |
7 | Germany | France | Netherlands | Switzerland |
8 | Italy | United Kingdom | Singapore | New Zealand |
9 | South Korea | Norway | India | Italy |
10 | New Zealand | South Korea | Turkey | Spain |
In essence, degree of centrality refers to the number of relational ties a node has in a network. In contrast, betweenness centrality refers to a node’s ability to connect otherwise unconnected groups of nodes, wherein nodes act as a gateway for the flow of information. Whereas, closeness centrality refers to a node’s closeness to every other node in the network, whereby nodes that reflect a greater number of shortest paths than others in a network indicates the ability of those nodes to transmit information and knowledge across the network with relative ease. Finally, eigen centrality refers to a node’s relative importance in a network, whereby nodes that are connected to other highly connected nodes are crucial to information transfer.
In terms of authors, Jian Mou emerged as the most important author for degree of centrality and betweenness centrality, whereas Xin Luo and Jian-xin Wang were flagged as the most important authors for closeness centrality and eigen centrality, respectively. In terms of institutions, Renmin University emerged as the most important institution for degree centrality and betweenness centrality, whereas the University of Ottawa was rated as the most important institution for closeness centrality and eigen centrality. In terms of countries, China emerged as the most important country for betweenness centrality, whereas the United States emerged as the most important country for the other three measures of centrality. Collectively, these findings indicate the most important constituents for degree of centrality, betweenness centrality, closeness centrality, and eigen centrality in terms of authors, institutions, and countries.
Collaboration network
The author collaboration network in Fig. 3 indicates that authors groups in ECR are fairly separated from each other, especially among highly connected authors (more than five links in the network). This suggests that most authors in ECR chose to work in a single team rather than across multiple teams. The institution collaboration network in Fig. 4 reaffirms our earlier finding that Renmin University is indeed the most important constituent of the network, especially among highly connected institutions (more than five links in the network). The institution collaboration network also appears to be more complex than the author collaboration network, wherein institutions appear to be far more connected to each other, indicating a good degree of collaboration across institutional lines. The country network in Fig. 5 presents a similar network scenario, where countries appear to be fairly well connected, with the United States being at the center of the country-level collaboration network. These findings suggest that ECR authors collaborate more actively across institutions and countries than teams.
Author co-authorship network. Note Threshold for inclusion is five or more links in the network
Institution co-authorship network. Note Threshold for inclusion is five or more links in the network
Country co-authorship network. Note Threshold for inclusion is five or more links in the network
Bibliographic coupling: thematic clusters
Bibliographic coupling is applied to unpack the major clusters (themes) within the ECR corpus. The method is predicated on the assumption that documents that share the same references are similar in content [ 18 , 35 ]. The application of bibliographic coupling on 443 ECR articles resulted in the formation of 30 clusters, wherein 11 major clusters were identified. The 11 major clusters, which contained 401 (or 90.5%) ECR articles, were ordered based on number of publications and average publication years, with more recent clusters ordered before older clusters in the case of clusters sharing the same number of publications. The summary of the 11 major clusters, which take center stage in this study, is presented in Table Table8 8 .
Summary of major clusters
Cluster # | Cluster theme | TP | TC | APY | Most cited articles | |||
---|---|---|---|---|---|---|---|---|
Author | Title | Year | TC | |||||
1 | Online privacy and security | 74 | 963 | 2013.09 | Zarmpou T., Saprikis V., Markos A., Vlachopoulou M | Modeling users’ acceptance of mobile services | 2012 | 121 |
Chaudhry S.A., Farash M.S., Naqvi H., Sher M | A secure and efficient authenticated encryption for electronic payment systems using elliptic curve cryptography | 2016 | 53 | |||||
Antoniou G., Batten L | E-commerce: protecting purchaser privacy to enforce trust | 2011 | 47 | |||||
2 | Online channels and optimization | 49 | 451 | 2016.67 | Jeffrey S.A., Hodge R | Factors influencing impulse buying during an online purchase | 2007 | 74 |
Biller S., Chan L.M.A., Simchi-Levi D., Swann J | Dynamic pricing and the direct-to-customer model in the automotive industry | 2005 | 62 | |||||
Yan R | Profit sharing and firm performance in the manufacturer-retailer dual-channel supply chain | 2008 | 43 | |||||
3 | Online engagement and preferences | 49 | 982 | 2013.98 | Nonnecke B., Andrews D., Preece J | Non-public and public online community participation: needs, attitudes and behavior | 2006 | 185 |
Sila I | Factors affecting the adoption of B2B e-commerce technologies | 2013 | 118 | |||||
Ozok A.A., Wei J | An empirical comparison of consumer usability preferences in online shopping using stationary and mobile devices: results from a college student population | 2010 | 69 | |||||
4 | Online market sentiments and analyses | 41 | 198 | 2018.56 | Zhou Q | Multi-layer affective computing model based on emotional psychology | 2018 | 63 |
Suki N.M | Consumer shopping behaviour on the Internet: insights from Malaysia | 2013 | 19 | |||||
Chen Y., Mullen T., Chu C.-H | An in-depth analysis of information markets with aggregate uncertainty | 2006 | 16 | |||||
5 | Online reviews and ratings | 40 | 611 | 2017.28 | Bae S., Lee T | Gender differences in consumers’ perception of online consumer reviews | 2011 | 125 |
Flanagin A.J., Metzger M.J., Pure R., Markov A., Hartsell E | Mitigating risk in ecommerce transactions: Perceptions of information credibility and the role of user-generated ratings in product quality and purchase intention | 2014 | 101 | |||||
Fairlie R.W | Explaining differences in access to home computers and the Internet: a comparison of Latino groups to other ethnic and racial groups | 2007 | 34 | |||||
6 | Online exchanges and transactions | 34 | 320 | 2011.29 | Narayanasamy K., Rasiah D., Tan T.M | The adoption and concerns of e-finance in Malaysia | 2011 | 39 |
Dumas M., Aldred L., Governatori G., Ter Hofstede A.H.M | Probabilistic automated bidding in multiple auctions | 2005 | 27 | |||||
Marinč M | Banks and information technology: Marketability vs. relationships | 2013 | 25 | |||||
7 | Online media and platforms | 30 | 668 | 2016.23 | Sotiriadis M.D., van Zyl C | Electronic word-of-mouth and online reviews in tourism services: the use of Twitter by tourists | 2013 | 188 |
Huang T.-L., Liao S | A model of acceptance of augmented-reality interactive technology: the moderating role of cognitive innovativeness | 2015 | 102 | |||||
Hsieh J.-K., Hsieh Y.-C., Tang Y.-C | Exploring the disseminating behaviors of eWOM marketing: persuasion in online video | 2012 | 80 | |||||
8 | Online technology acceptance and continuance | 26 | 249 | 2016.37 | Zhou T | An empirical examination of user adoption of location-based services | 2013 | 42 |
Chen Q., Chen H.-M., Kazman R | Investigating antecedents of technology acceptance of initial ECRM users beyond generation X and the role of self-construal | 2007 | 34 | |||||
Royo S., Yetano A | “Crowdsourcing” as a tool for e-participation: two experiences regarding CO2 emissions at municipal level | 2015 | 22 | |||||
9 | Online communities and commercialization in the virtual world | 22 | 771 | 2012.23 | Füller J., Bartl M., Ernst H., Mühlbacher H | Community based innovation: How to integrate members of virtual communities into new product development | 2006 | 270 |
Lehdonvirta V | Virtual item sales as a revenue model: Identifying attributes that drive purchase decisions | 2009 | 170 | |||||
Guo Y., Barnes S | Virtual item purchase behavior in virtual worlds: An exploratory investigation | 2009 | 97 | |||||
10 | Online customer expectations, satisfaction, and loyalty | 18 | 291 | 2016.11 | Hanafizadeh P., Khedmatgozar H.R | The mediating role of the dimensions of the perceived risk in the effect of customers’ awareness on the adoption of Internet banking in Iran | 2012 | 63 |
Valvi A.C., Fragkos K.C | Critical review of the e-loyalty literature: a purchase-centred framework | 2012 | 60 | |||||
Aloudat A., Michael K | Toward the regulation of ubiquitous mobile government: a case study on location-based emergency services in Australia | 2011 | 39 | |||||
11 | Online purchase intention | 18 | 671 | 2014.00 | Kim J.B | An empirical study on consumer first purchase intention in online shopping: integrating initial trust and TAM | 2012 | 124 |
Gregg D.G., Walczak S | The relationship between website quality, trust and price premiums at online auctions | 2010 | 104 | |||||
Taylor D.G., Davis D.F., Jillapalli R | Privacy concern and online personalization: the moderating effects of information control and compensation | 2009 | 76 |
Cluster #1: online privacy and security
Cluster #1 contains 74 articles that have been cited 963 times with an average publication year of 2013.09. The most cited article in this cluster is Zarmpou et al.’s [ 36 ] article on the adoption of mobile services. This is followed by Chaudhry et al.’s [ 37 ] article on user encryption schemes for e-payment systems, and Antoniou and Batten’s [ 38 ] article on purchaser’s privacy and trust in online transactions. Other articles in this cluster have considered topics such as e-commerce trust models [ 39 ], consumer privacy [ 40 ], cybercrime and cybersecurity issues [ 41 ], gender differences [ 42 ], and the development and implementation of various authentication systems [ 43 , 44 ]. Thus, ECR articles in this cluster appear to be centered on online privacy and security issues , including equivalent solutions for improved authentication and encryption to improve trust in electronic commerce.
Cluster #2: online channels and optimization
Cluster #2 contains 49 articles that have been cited 415 times with an average publication year of 2016.67. The most cited article in this cluster is Jeffrey and Hodge’s [ 45 ] article on impulse purchases in online shopping. This is followed by Biller et al.’s [ 46 ] article on dynamic pricing for online retailing in the automotive industry, and Yan’s [ 47 ] article on profit sharing and firm performance in manufacturer-retailer dual-channel supply chains. Other articles in this cluster have examined online channels such as peer-to-peer networks and social commerce [ 48 , 49 ] and optimal supply chain configuration [ 50 , 51 ]. Thus, ECR articles in this cluster appear to be concentrated on online channels and optimization , particularly in terms of the channel characteristics and price and supply chain optimization in electronic commerce.
Cluster #3: online engagement and preferences
Cluster #3 contains 49 articles that have been cited 982 times with an average publication year of 2013.98. The most cited article in this cluster is Nonnecke et al.’s [ 28 ] article on online community participation. This is followed by Sila’s [ 52 ] article on business-to-business electronic commerce technologies, and Ozok and Wei’s [ 53 ] article on consumer preferences of using mobile and stationary devices. Other articles in this cluster have explored topics such as online community participation and social impact across countries [ 54 ], online opinions across regions and its impact on consumer preferences [ 55 , 56 ], content and context factors [ 57 ], data mining techniques [ 58 ], and recommender systems and their application in online environments [ 59 , 60 ]. Thus, ECR articles in this cluster appear to be focused on online engagement and preferences , including the adoption and usage of technology (e.g., data mining, recommender systems) to curate engagement and shape preferences among target customers in electronic commerce.
Cluster #4: online market sentiments and analyses
Cluster #4 contains 41 articles that have been cited 198 times. This cluster has the highest average publication year among the 11 major clusters (2018.56), which indicates that most articles in this cluster are fairly recent. The most cited article in this cluster is Zhou’s [ 61 ] article on multi-layer affective modeling of emotions in the online environment. This is followed by Suki’s [ 62 ] article on online consumer shopping insights, and Chen et al.’s [ 63 ] article on information markets. Other articles in this cluster have investigated topics such as Internet queries and marketplace prediction [ 64 ], cross-border electronic commerce using the information systems success model [ 65 ], and electronic [ 66 ] and social [ 67 ] commerce using big data. Thus, ECR articles in this cluster appear to be centered on online market sentiments and analyses , with the use of advanced modeling techniques to unpack fresh insights on electronic commerce being relatively prominent.
Cluster #5: online reviews and ratings
Cluster #5 contains 40 articles that have been cited 611 times with an average publication year of 2017.28. The most cited article in this cluster is Bae and Lee’s [ 30 ] article on online consumer reviews across gender. This is followed by Flanagin et al.’s [ 68 ] article on user-generated online ratings, and Fairlie’s [ 69 ] on the digital divide in online access, which speaks to the technological infrastructure required to post and respond to online reviews and ratings. Other articles in this cluster have examined quantitative and qualitative feedback in online environments [ 70 ], electronic word of mouth platforms and persuasiveness [ 71 ], online reviews and product innovation [ 72 ] , recommender systems and product ranking [ 73 ], and online rating determinants [ 74 ]. Thus, ECR articles in this cluster appear to be concentrated on online reviews and ratings , including its potential differences among consumers coming from different demographic backgrounds.
Cluster #6: online exchanges and transactions
Cluster #6 contains 34 articles that have been cited 320 times with an average publication year of 2011.29. The most cited article in this cluster is Narayanasamy et al.’s [ 75 ] article on the adoption and concerns of e-finance. This is followed by Dumas et al.’s [ 76 ] article on bidding agents in e-auction, and Marinč’s [ 77 ] article on the impact of information technology on the banking industry. Other articles in this cluster have explored topics such as game theoretic aspects of search auctions [ 78 ], auction mechanism for ad space among advertisers [ 79 ], trust analysis in online procurement [ 80 ], efficiency of reverse auctions [ 81 ], and effect of hedonic and utilitarian behaviors on the e-auction behavior [ 82 ]. Thus, ECR articles in this cluster appear to be focused on online exchanges and transactions , particularly in terms of auction mechanisms and banking-related services.
Cluster #7: online media and platforms
Cluster #7 contains 30 articles that have been cited 668 times with an average publication year of 2016.23. The most cited article in this cluster is Sotiriadis and van Zyl’s [ 27 ] article on social media in the form of Twitter. This is followed by Huang and Liao’s [ 83 ] article on augmented reality interactive technology, and Hsieh et al.’s [ 84 ] article on online video persuasion in electronic commerce. Other articles in this cluster have investigated topics such as the role of social media in disseminating product information [ 85 ], the effect of video formats on person-to-person streaming [ 86 ], interpersonal relationship building using social media [ 87 ], and microblog usage [ 88 ]. Thus, ECR articles in this cluster appear to be centered on online media and platforms , particularly in terms of its variation, use, and impact in shaping consumer behavior in electronic commerce.
Cluster #8: online technology acceptance and continuance
Cluster #8 contains 26 articles that have been cited 244 times with an average publication year of 2016.37. The most cited article in this cluster is Zhou’s [ 89 ] article on the adoption of location-based services. This is followed by Chen et al.’s [ 90 ] article on the adoption of electronic customer relationship management, and Royo and Yetano’s [ 91 ] article on crowdsourcing usage in local governments. Other articles in this cluster have examined topics such as gender discrimination in online peer-to-peer lending [ 92 ], continued usage of e-auction services [ 93 ], and investor trust in peer-to-peer lending platforms [ 94 ]. Thus, ECR articles in this cluster appear to be concentrated on online technology acceptance and continuance , including determinants and discriminants that explain online technology-mediated behavior across different forms of electronic commerce such as e-auction, e-lending, e-government, and e-customer relationship management.
Cluster #9: online communities and commercialization in the virtual world
Cluster #9 contains 22 articles that have been cited 771 times with an average publication year of 2012.23. The most cited article in this cluster is Füller et al.’s [ 26 ] article on the role of virtual communities in new product development. This is followed by Lehdonvirta’s [ 29 ] article on the revenue model of virtual products, and Guo and Barnes’s [ 95 ] article on the purchase behavior of virtual products. Other articles in this cluster have investigated topics such as metaverse retailing [ 96 ], issues faced by developers of virtual worlds [ 97 ], the impact of virtual world on e-business models [ 98 ], e-commerce transactions in virtual environments [ 99 ], and customer value co-creation in virtual environments [ 26 ]. Thus, ECR articles in this cluster appear to be focused on the online communities and commercialization in the virtual world , particularly in virtual environments such as online gaming.
Cluster #10: online customer expectations, satisfaction, and loyalty
Cluster #10 contains 18 articles that have been cited 291 times with an average publication year of 2016.11. The most cited article in this cluster is Hanafizadeh and Khedmatgozar’s [ 100 ] article on consumer expectations of risk in online banking. This is followed by Valvi and Fragkos’s [ 101 ] article on purchase-centered e-loyalty, and Aloudat and Michael’s [ 102 ] article on regulatory expectations of ubiquitous mobile government. Other articles in this cluster have examined topics such as continued usage of e-services [ 103 ], determinants of e-loyalty [ 104 ] , risk expectations of e-services [ 105 ], and e-service quality implications for customer satisfaction and loyalty [ 106 ]. Thus, ECR articles in this cluster appear to be centered on online customer expectations, satisfaction, and loyalty , particularly in e-service settings such as online banking.
Cluster #11: online purchase intention
Cluster #11 contains 18 articles that have been cited 671 times with an average publication year of 2014.00. The most cited article in this cluster is Kim’s [ 107 ] article on online purchase intention using trust theory and technology acceptance model. This is followed by Gregg and Walczak’s [ 108 ] article on the effects of website quality on online purchase intention, and Taylor et al.’s [ 109 ] article on the effects of privacy concerns on online purchase intention. Other articles in this cluster have explored topics that either reaffirm the findings of the highly cited articles in this cluster, such as privacy concerns and personalization [ 109 , 110 ], or that extend the breadth of cluster coverage, such as store image [ 111 ], risk, and trust [ 112 ] as determinants of online purchase intention. Thus, ECR articles in this cluster appear to be concentrated on online purchase intentions , particularly in terms of its multi-faceted determinants that avail or transpire in electronic commerce.
Temporal keyword analysis: thematic evolution
Building on the thematic clusters uncovered using bibliographic coupling (see Fig. 6 ), this study performs a temporal keyword analysis to unpack the development of themes and its evolutionary trajectory in ECR over time.
Period wise publication trend in major clusters. Note Cluster #1 = online privacy and security. Cluster #2 = online channels and optimization. Cluster #3 = online engagement and preferences. Cluster #4 = online market sentiments and analyses. Cluster #5 = online reviews and ratings. Cluster #6 = online exchanges and transactions. Cluster #7 = online media and platforms. Cluster #8 = online technology acceptance and continuance. Cluster #9 = online communities and commercialization in the virtual world. Cluster #10 = online customer expectations, satisfaction, and loyalty. Cluster #11 = online purchase intention
Thematic development from 2005 to 2008
Most ECR articles between 2005 and 2008 appear in Clusters #1, #3, and #6 (see Fig. 6 ), which indicate research concentration on online privacy and security, online engagement and preferences, and online exchanges and transactions. The keyword network in Fig. 7 confirms this observation. Apart from general keywords such as “e-commerce,” keywords such as “cryptography,” “privacy,” and “security” relate directly to the theme of Cluster #1, which is about online privacy and security. The prominence of the word “cryptography” indicates the popularity and importance of the topic during this period. Other keywords such as “auctions,” “online auctions,” and “bidding strategies” relate to the theme of Cluster #6, which is about online exchanges and transactions, with particular focus on online auction and banking. Other keywords such as “collaborative filtering,” “online communities,” and “mobile commerce” relate to the theme of Cluster #3, which is about online engagement and preferences. The bigger and bolder keywords observed in Clusters #1 and #3 suggest that the direct benefits and costs of electronic commerce were most pertinent in the early stages of ECR , with the augmented aspects of electronic commerce in Cluster #6 emerging closely behind the two leading clusters in this period.
Keyword network between 2005 and 2008. Note Threshold for inclusion is a minimum of two occurrences
Thematic development from 2009 to 2012
Most ECR articles between 2009 and 2012 are located in Cluster #1 (see Fig. 6 ), which reveal the continued pertinence of research concentrating on online privacy and security during this period. Nonetheless, ECR experienced a substantial growth in research focusing on online media and platforms, online communities and commercialization in the virtual world, online customer expectations, satisfaction, and loyalty, and online purchase intention, as seen through ECR articles in Clusters #7, #9, #10, and #11 during this period. The keyword network in Fig. 8 adds to this observation. In particular, keywords such as “security,” “payment protocol,” and “trust management” relate to the theme of Cluster #1 on online privacy and security, whereas keywords such as “metaverses,” “second life,” “virtual reality,” and “virtual world” speak to the emergence of online communities and commercialization in the virtual world characterizing Cluster #9. Similarly, keywords such as “reputation” and “trust” are important to online customer expectations, satisfaction, and loyalty (Cluster #10) and their online purchase intention (Cluster #11). Interestingly, though Cluster #7 emerged during this period, we did not observe any unique or specific keywords relating to this cluster, which may be attributed to online media and platform research early focus on its “adoption,” a keyword that we felt resonates more with Cluster #8.
Keyword network between 2009 and 2012. Note Threshold for inclusion is a minimum of two occurrences
Thematic development from 2013 to 2016
Most ECR articles between 2013 and 2016 continue to be situated in Cluster #1 (see Fig. 6 ), which suggest the continued pertinence of research concentrating on online privacy and security during this period. Nonetheless, there are a number of clusters that saw noteworthy growth, such as Clusters #2, #5, #7, #8, and #10, which indicate that research attention has also been invested in topics related to online channels and optimization, online reviews and ratings, online media and platforms, online technology acceptance and continuance, and online customer expectations, satisfaction, and loyalty. The keyword network in Fig. 9 supports this observation. More specifically, keywords such as “personal information” and “privacy” indicate continued research in Cluster #1, though it appears that the focus has shifted from authentication and security mechanisms to privacy matters, which may be attributed to the rise of personalized and targeted online marketing activities (e.g., tracking of user activity for personalized advertisements). Whereas, keywords such as “B2C e-commerce” and “e-government” denote emerging interest in online channels and optimization (Cluster #2), “electronic word of mouth” indicates growing interest in online reviews and ratings (Cluster #5), “cloud computing,” “IPTV,” and “social media” reveal increasing interest in online media and platforms (Cluster #7), “information technology,” “technology adoption,” and “technology acceptance model” speak to research on online technology acceptance and continuance (Cluster #8), and “product type,” “quality of service,” and “user satisfaction” resonate with research on online customer expectations, satisfaction, and loyalty (Cluster #10).
Keyword network between 2013 and 2016. Note Threshold for inclusion is a minimum of two occurrences
Thematic development from 2017 to 2020
Most ECR articles between 2017 and 2020 are located in Cluster #4 (see Fig. 6 ), which reflect the noteworthy emergence and shift of research concentration from online privacy and security to online market sentiments and analyses. Other thematic clusters such as Clusters #2, #3, and #5 have also witnessed a massive increase in publications during this period. This implies that ECR has become relatively diverse in the research that it publishes, which also explains the rise in the number of papers that the journal publishes during this period. The keyword network in Fig. 10 sheds further light on this observation. In particular, many keywords in the network illustrate a strong research concentration on online market sentiments and analyses, such as “big data,” “data mining,” machine learning,” “sentiment analysis,” and “social network analysis” (Cluster #4). Similarly, keywords such as “dual channel supply chain,” “supply chain coordination,” and “social commerce” indicate the type of research focusing on online channels and optimization (Cluster #2), “social influence,” “social media,” and “social media marketing” reflect research in the area of online engagement and preferences (Cluster #3), and “consumer reviews,” “online reviews,” “reputation,” and “word of mouth” speak to research on online reviews and ratings (Cluster #5).
Keyword network between 2017 and 2020. Note Threshold for inclusion is a minimum of two occurrences
This study presents a 20-year retrospective of ECR since its inception in 2001. Several research questions were proposed and pursued using a bibliometric methodology consisting of performance analysis and science mapping (e.g., co-authorship analysis, bibliographic coupling, and temporal keyword analysis).
Our first four research questions—i.e., research question 1 to research question 4—concentrated on the publication and citation trends of ECR . Through performance analysis, we found that ECR has grown exponentially in terms of its publications and citations. Most contributors of ECR come from China and the United States, which reflect (1) China’s standing as the world’s largest e-commerce market with 50 percent of the world’s online transactions occurring in this country, and (2) the United States’ standing as the world’s pioneer of e-commerce (e.g., Amazon) and her expectation for e-commerce to reach 50% of total retail sales in the country in 10 years [ 113 ]. Interestingly, IBM, a non-academic institution, emerged as the highest contributing institution to the journal, which is unsurprising given that IBM is the largest industrial research organization in the world with 12 research labs across six continents [ 114 ]. More importantly, ECR was found to be well received among its peers, with many of its citations coming from prestigious journals in the field of information systems and management. Nevertheless, we observed that ECR receives very little contribution from Africa and several parts of Asia, particularly South Asia and South East Asia. Though electronic commerce may not have been very prominent in these regions in the past, we believe that the coronavirus pandemic that has taken the world by storm in 2020 has accelerated the proliferation and adoption of electronic commerce in these regions, and thus, we would encourage authors from these regions to submit their best papers to ECR in the near future. Thus, we raise two future research questions (FRQs) for exploration:
FRQ1: What are the e-commerce innovations that avail in underexplored regions (e.g., Africa, South Asia, and South East Asia) and how do such innovations fare in terms of similarities and differences in manifestations and impact against their more richly explored counterparts (e.g., China, United States)?
FRQ2: How can global pandemics such as COVID-19 change or impact e-commerce around the world (e.g., can the pandemic accelerate e-commerce adoption across all layers of society; can the pandemic lead to new innovations; can e-commerce contribute to positive and/or negative economic and social impact during the pandemic—and if yes, what and how, and if no, why)?
Our next two research questions—i.e., research question 5 and research question 6—focused on the collaboration trends in and the important constituents of ECR in the co-authorship network. Using co-authorship analysis, we found that the collaboration culture in ECR has grown with the passage of time, as evidenced through the decreasing share of single-authored articles and the increasing share of multi-authored publications, especially in the five or more authors category. We also observed that the share of multi-authored articles has always been dominant in the journal, with such publications forming nearly 90% of the corpus at any given point in time. Indeed, these observations reflect the increasing emphasis that universities place on multi-author and inter-/multi-/trans-disciplinary collaborations in promotion and tenure practices and policies [ 115 ]. In terms of important constituents in the co-authorship network, Jian Mou emerged as the most important author across two measures of centrality, whereas Renmin University and University of Ottawa emerged as the most important institutions at the institution level, and the United States emerged as the most important constituent at the country level. Nonetheless, we noted that authors who collaborate in ECR do not work much across diverse teams, but they do, however, work a lot across institutions and countries. Future scholars could rely on the centrality networks that we have curated herein this study for potential collaboration with authors from varying institutions and countries who have a good publication record and a research interest to publish with ECR .
Our final research question—i.e., research question 7—was dedicated to unpacking the major themes in ECR . Through bibliographic coupling, our study found 11 major clusters that reflected the major themes underpinning research published in ECR : (1) online privacy and security, (2) online channels and optimization, (3) online engagement and preferences, (4) online market sentiments and analyses, (5) online reviews and ratings, (6) online exchanges and transactions, (7) online media and platforms, (8) online technology acceptance and continuance, (9) online communities and commercialization in the virtual world, (10) online customer expectations, satisfaction, and loyalty, and (11) online purchase intention. Through temporal keyword analysis, our study observed that the topics published in ECR has become more diverse over time, with a noteworthy shift from an early concentration on online privacy and security to a contemporary focus on newer, industry-informed topics, such as online market sentiments and analyses, which we reckon coincides with the emergence of the unique peculiarities of the fourth industrial revolution (IR 4.0), such as big data and machine learning, in recent years [ 116 , 117 ]. Thus, to extend the line of research that concentrates on unpacking the contemporary realities of e-commerce, we propose another two future research questions (FRQs) for exploration:
FRQ3: How can emergent technologies (e.g., artificial intelligence, big data analytics, blockchain, machine learning) be applied to improve forecasting (e.g., cybercrime, social network), optimize functions (e.g., advertising, sales), and protect stakeholders (e.g., privacy, security) in e-commerce?
FRQ4: How can e-commerce operators leverage on emergent technologies to acquire competitive advantages (e.g., how to build trust and good relationships with customers [e.g., digital natives, digital migrants], and how to respond to changes in customer demands and marketplace trends with agility), and whether these competitive advantages that they acquired are sustainable or transient (and if transient, then what can they do to curate, maintain, or replenish their competitive advantages in the long run)?
Though thorough in its approach, this study does suffer from certain limitations. First, this study relies on the Scopus for bibliometric data. Though the database has its merits, as laid out in the methodology section, the bibliographic data is not created for the purpose of bibliometric analysis. This may lead to errors in the data source. Through data cleaning, we have attempted to minimize errors, but any remaining error in the source data, which we might have missed, could have an impact on the final analysis, though we believe that the margin for such errors would be relatively small, if not, negligible. Second, ECR has been around for 20 years, but the dataset available on Scopus, which we used, is only complete for 16 years (2005–2020). Due to this limitation, the science mapping part of the study—i.e., co-authorship, bibliographic coupling, and temporal keyword analysis—had to be restricted to this period only. We do not discount the possibility that the complete set of earlier data (2001–2004) may become available on Scopus in the future, and thus, we would encourage future research aiming to conduct a bibliometric review for ECR , perhaps in the next milestone (e.g., 30, 40, or 50 years), to check on such data availability, and if available, to take advantage and conduct a full-fledged science mapping for the journal. Finally, the scientific insights that could be uncovered through a bibliometric methodology, though rich, remain limited. In particular, bibliometric reviews such as ours do not delve into expert information, such as the theories, contexts, and methods employed to create new knowledge on electronic commerce in the ECR corpus. This, in turn, makes it difficult for bibliometric reviews to put forth a comprehensive set of data-informed proposals for future research. Nonetheless, we opine that bibliometric reviews do provide a good starting point of data-informed insights that future research can rely on to understand the trajectory of the extant discussion of electronic commerce in the journal. In particular, we believe that such insights would be useful, not only for future empirical research (e.g., potential collaboration networks, research themes of interest), but also for future reviews on thematic domains in ECR (e.g., systematic reviews on online market sentiments), which can be done in a number of ways, such a critical review [ 118 – 120 ], a thematic review [ 121 , 122 ], a theory-driven review [ 123 ], a method-driven review [ 124 , 125 ], or a framework-based review [ 126 ].
Compliance with ethical standards
On behalf of all authors, the corresponding author states that there is no conflict of interest.
1 Web of Science single-year and five-year impact factors for ECR : https://www.springer.com/journal/10660 .
2 Scopus CiteScore and SNIP for ECR : https://www.scopus.com/sourceid/145669 .
Contributor Information
Satish Kumar, Email: [email protected] .
Weng Marc Lim, Email: moc.cramgnew@mil , Email: ua.ude.niws@milcram , Email: ym.ude.enrubniws@milw .
Nitesh Pandey, Email: ni.ca.tinm@6109MBR8102 .
J. Christopher Westland, Email: ude.ciu@dnaltsew .
Information
- Author Services
Initiatives
You are accessing a machine-readable page. In order to be human-readable, please install an RSS reader.
All articles published by MDPI are made immediately available worldwide under an open access license. No special permission is required to reuse all or part of the article published by MDPI, including figures and tables. For articles published under an open access Creative Common CC BY license, any part of the article may be reused without permission provided that the original article is clearly cited. For more information, please refer to https://www.mdpi.com/openaccess .
Feature papers represent the most advanced research with significant potential for high impact in the field. A Feature Paper should be a substantial original Article that involves several techniques or approaches, provides an outlook for future research directions and describes possible research applications.
Feature papers are submitted upon individual invitation or recommendation by the scientific editors and must receive positive feedback from the reviewers.
Editor’s Choice articles are based on recommendations by the scientific editors of MDPI journals from around the world. Editors select a small number of articles recently published in the journal that they believe will be particularly interesting to readers, or important in the respective research area. The aim is to provide a snapshot of some of the most exciting work published in the various research areas of the journal.
Original Submission Date Received: .
- Active Journals
- Find a Journal
- Proceedings Series
- For Authors
- For Reviewers
- For Editors
- For Librarians
- For Publishers
- For Societies
- For Conference Organizers
- Open Access Policy
- Institutional Open Access Program
- Special Issues Guidelines
- Editorial Process
- Research and Publication Ethics
- Article Processing Charges
- Testimonials
- Preprints.org
- SciProfiles
- Encyclopedia
Journal Menu
- Aims & Scope
- Editorial Board
- Instructions for Authors
- Special Issues
- Sections & Collections
- Article Processing Charge
- Indexing & Archiving
- Most Cited & Viewed
- Journal Statistics
- Journal History
- Society Collaborations
- Editorial Office
Journal Browser
- arrow_forward_ios Forthcoming issue arrow_forward_ios Current issue
- Vol. 19 (2024)
- Vol. 18 (2023)
- Vol. 17 (2022)
- Vol. 16 (2021)
- Volumes not published by MDPI
- Vol. 15 (2020)
- Vol. 14 (2019)
- Vol. 13 (2018)
- Vol. 12 (2017)
- Vol. 11 (2016)
- Vol. 10 (2015)
- Vol. 9 (2014)
- Vol. 8 (2013)
- Vol. 7 (2012)
- Vol. 6 (2011)
- Vol. 5 (2010)
- Vol. 4 (2009)
- Vol. 3 (2008)
- Vol. 2 (2007)
- Vol. 1 (2006)
Find support for a specific problem in the support section of our website.
Please let us know what you think of our products and services.
Visit our dedicated information section to learn more about MDPI.
Open Access and Article Processing Charge (APC)
All articles published in Journal of Theoretical and Applied Electronic Commerce Research (ISSN 0718-1876) are published in full open access . An article processing charge (APC) of CHF 1000 (Swiss Francs) applies to papers accepted after peer review. This article processing charge is to cover the costs of peer review, copyediting, typesetting, long-term archiving, and journal management. In addition to Swiss francs (CHF), we also accept payment in euros (EUR), US dollars (USD), British pound sterling (GBP), Japanese yen (JPY) or Canadian dollars (CAD).
Invoices are emailed shortly after acceptance to the payment contact provided by the authors. Only official invoices issued by MDPI (@mdpi.com) are valid. We do not authorize any third party to collect the APCs. MDPI is the sole service provider and cannot be held liable for actions by third parties.
Local VAT or Sales Tax will be added if applicable.
For more detailed information on how to pay an invoice, please see our Payment Instructions .
Submitted papers should be well formatted and use good English. Authors may use MDPI's English editing service prior to publication or during author revisions. Note that many national and private research funding organizations and universities explicitly cover APCs for articles resulting from funded research projects. Discounts are also available for authors from institutes which participate in MDPI's Institutional Open Access Program ( IOAP ).
- Article Processing Charges (APC) Information
- Open Access Information and Policy
- Information for Authors
- Accepted Methods of Payment
- Pay an Invoice Now
- MDPI Invoicing FAQ
Further Information
Mdpi initiatives, follow mdpi.
Subscribe to receive issue release notifications and newsletters from MDPI journals
- DOI: 10.1007/978-3-319-50091-1
- Corpus ID: 168662352
- Introduction to Electronic Commerce and Social Commerce
- E. Turban , J. Whiteside , +1 author Jon Outland
- Published 4 May 2017
- Computer Science, Business
68 Citations
E-commerce application (q-kart), an overview of electronic commerce (e-commerce), an analysis of the impact, problems, and challenges of digital platforms for e-commerce marketing, marketing and technologies platforms in smart f-store, igniting social commerce: using instagram for mobile retail shopping.
- Highly Influenced
Understanding global e-commerce development during the COVID-19 pandemic: Technology-Organization-Environment perspective
A bibliometric analysis of platform research in e-commerce: past, present, and future research agenda, critical factors in indonesia's e-commerce collaboration, factors influencing the purchase decision through social commerce channels: the results of mixed research in moscow, instagram shopping in saudi arabia: what influences consumer trust and purchase decisions, related papers.
Showing 1 through 3 of 0 Related Papers
IEEE Account
- Change Username/Password
- Update Address
Purchase Details
- Payment Options
- Order History
- View Purchased Documents
Profile Information
- Communications Preferences
- Profession and Education
- Technical Interests
- US & Canada: +1 800 678 4333
- Worldwide: +1 732 981 0060
- Contact & Support
- About IEEE Xplore
- Accessibility
- Terms of Use
- Nondiscrimination Policy
- Privacy & Opting Out of Cookies
A not-for-profit organization, IEEE is the world's largest technical professional organization dedicated to advancing technology for the benefit of humanity. © Copyright 2024 IEEE - All rights reserved. Use of this web site signifies your agreement to the terms and conditions.
Journal of Electronic Commerce Research
Regular Issue
- Consumer Responses to Online Advertising and Online Reviews: The Mediating Role of Cognitive Responses and Affective Responses
- Do Mhealth Apps Influence Consumers’ Safe Food Choice Decisions? Role of Technology Affordances
- Promotion Regulatory Fit or Prevention Regulatory Fit, Which Is the King on Social Media? Influence of Social Identification on Brand Loyalty
- Attention to Product Images in an Online Retailing Store: An Eye-Tracking Study Considering Consumer Goals and Type of Product
Most Viewed Articles
- Content Analysis Of Social Media: A Grounded Theory Approach
- Customer Willingness To Pay For Online Music: The Role Of Free Mentality
- Pre-Purchase Online Information Seeking: Search Versus Browse
- Consumer Responses To Brands Placed In Youtube Movies: The Effect Of Prominence And Endorser Expertise
- Optimal Reserve Price For The Generalized Second-Price Auction In Sponsored Search Advertising
- Providing Information And Enabling Transactions: Which Website Function Is More Important For Success?
Recommended Articles
- Investigating Collaborative Commerce System from the Perspective of Collaborative Relationship
- Cultural Adaptation In Cross Border E-Commerce: A Study Of German Companies
- Investigating Common Method Bias via an EEG Study of the Flow Experience in Website Design
Investigating LLM Applications in E-Commerce
The emergence of Large Language Models (LLMs) has revolutionized natural language processing in various applications especially in e-commerce. One crucial step before the application of such LLMs in these fields is to understand and compare the performance in different use cases in such tasks. This paper explored the efficacy of LLMs in the e-commerce domain, focusing on instruction-tuning an open source LLM model with public e-commerce datasets of varying sizes and comparing the performance with the conventional models prevalent in industrial applications. We conducted a comprehensive comparison between LLMs and traditional pre-trained language models across specific tasks intrinsic to the e-commerce domain, namely classification, generation, summarization, and named entity recognition (NER). Furthermore, we examined the effectiveness of the current niche industrial application of very large LLM, using in-context learning, in e-commerce specific tasks. Our findings indicate that few-shot inference with very large LLMs often does not outperform fine-tuning smaller pre-trained models, underscoring the importance of task-specific model optimization.Additionally, we investigated different training methodologies such as single-task training, mixed-task training, and LoRA merging both within domain/tasks and between different tasks. Through rigorous experimentation and analysis, this paper offers valuable insights into the potential effectiveness of LLMs to advance natural language processing capabilities within the e-commerce industry.
1. Introduction
Large Language Models (LLMs) have recently gained ubiquitous use in many domains (Singhal et al . , 2022 ; Trautmann et al . , 2022 ; Malinka et al . , 2023 ) . In the e-commerce domain in particular, LLMs have the potential to facilitate the creation of product descriptions, summarize reviews, expand queries, and answer buyer and seller questions, among other potential use cases.
We simulated a common practice in the e-commerce industry: adapting open source state of the art models for domain specific tasks. We compared the feasibility of using an LLM and explored to what extent using an LLM for different e-commerce tasks leads to gains in the evaluation metrics. Training LLMs from scratch requires a significant amount of resources, but a common practice to efficiently train the model is to use parameter efficient methods like low rank adapters (LoRA) (Hu et al . , 2021 ) . Important questions arise when attempting to adapt these models for specific tasks and domains. We attempt to answer in this work: How much training data is needed to adapt a model to a task? How much do LLMs outperform more traditional approaches? What ways do different tasks interact with each other when doing mixed dataset training or merging LoRA weights trained on indivudal tasks? There are various approaches for adapting LLMs for tasks in a specific domain. Specifically, we focus on LoRA supervised fine-tuning (SFT), multi-task training, zero-shot inference, and LoRA merging.
Our contributions are as follows: 1) Organizing and formatting four e-commerce datasets for the evaluation of large language models (LLMs). 2) Conducting comprehensive experiments to compare fine-tuning a large language model with conventional industrial baselines, such as BERT and T5, using varying amounts of data for e-commerce tasks; additionally, we examine the effectiveness of in-context learning with a highly competitive, very large LLM. 3) Exploring the use of mixed LoRA merging across different tasks and comparing this approach to traditional mixed dataset training.
Our findings indicate that for e-commerce-specific tasks, conventional methods, such as training smaller language models, can achieve performance that is comparable to or even surpasses that of general-purpose very large LLMs. These valuable insights provide for the application of these models within the e-commerce industry.
2. Background
Large Language Models (LLMs) (Chowdhery et al . , 2023 ; OpenAI et al . , 2023 ; Anil et al . , 2023 ; Almazrouei et al . , 2023 ; Touvron et al . , 2023 ) have seen increasing attention in recent years as models that perform natural language generation have begun to be used for multiple tasks. They differ from prior pre-trained language models (PLMs), such as BERT (Devlin et al . , 2019 ) or T5 (Raffel et al . , 2020 ) , in their amount of training data and number of parameters.
2.1. Instruction Fine-tuning
Instruction fine-tuning represented a pivotal advancement in the optimization of large language models (LLMs), such as GPT-4, for enhanced task-specific performance, especially in domain-specific applications (Hu et al . , 2023 ; Zhang et al . , 2024 ) . This method involved the supplementary training of a pre-trained based model such as GPT (OpenAI et al . , 2023 ) , Llama (Touvron et al . , 2023 ) , or Falcon (Almazrouei et al . , 2023 ) on a task specific dataset consisting of prompts paired with their optimal responses. The objective was to refine the model’s capacity to comprehend and execute instructions with increased accuracy and contextual relevance. Instruction fine-tuning has emerged as an invaluable technique for augmenting the proficiency of LLMs across various specialized domains, ensuring their outputs align more closely with user expectations and requirements.
2.2. Low-Rank Adaptation Training
Low-Rank Adaptation (LoRA) (Hu et al . , 2021 ) is an innovative technique designed to fine-tune (LLMs) in a resource-efficient manner. This method addresses the challenge of adapting pre-trained models to specialized tasks without the extensive computational costs associated with traditional full-model fine-tuning. At the heart of LoRA is the strategic introduction of trainable low-rank matrices that target specific components of the LLM’s architecture, namely the attention and feed-forward neural network layers inherent to the transformer model. Specifically, it freezes the pre-trained layers of the LLM, and for each layer, it trains a rank-decomposition matrix and injects them into each layer of the pre-trained model to accomplish the LLM fine-tuning.
𝐖 𝐀 superscript 𝐁 𝑇 \mathbf{W}^{\prime}=\mathbf{W}+\mathbf{A}*\mathbf{B}^{T} bold_W start_POSTSUPERSCRIPT ′ end_POSTSUPERSCRIPT = bold_W + bold_A ∗ bold_B start_POSTSUPERSCRIPT italic_T end_POSTSUPERSCRIPT , where 𝐖 ′ superscript 𝐖 ′ \mathbf{W}^{\prime} bold_W start_POSTSUPERSCRIPT ′ end_POSTSUPERSCRIPT represents the adapted weights. This process selectively fine-tunes the model, allowing it to acquire new capabilities or improve performance on specific tasks with minimal adjustments to its pre-trained parameters. This selective updating is particularly beneficial for domain-specific applications, where only certain aspects of the model’s knowledge need refinement.
2.3. Evaluation
With the rise of text generation models that are seemingly capable of performing large numbers of tasks and able to answer many questions, a number of evaluation strategies have been proposed. Evaluation leaderboards often consist of evaluation tasks like Hellaswag (Zellers et al . , 2019 ) , MMLU (Hendrycks et al . , 2021b ) , and others (Gao et al . , 2023 ; Hendrycks et al . , 2021b , a ) which cover a broad range of multiple choice questions. These types of multiple choice question evaluations, where the answer is chosen based on the choice with the highest likelihood can be brittle as rankings can be sensitive to minute details (Alzahrani et al . , 2024 ) . Other evaluation benchmarks like GLUE (Wang et al . , 2018 ) consist of a bundle of different tasks with task specific evaluation metrics. Another approach to evaluating LLM performance is the approach of LLM as a judge. Chatbot arena (Zheng et al . , 2024 ) is an example of this style of evaluation and while it can correlate with human judgment, like other LLM applications it can be subject to hallucinations. In this work, we focus on directly evaluating the tasks of interest with existing scoring practices.
2.4. LLMs for E-commerce
Zhang et al . ( 2024 ) find that fine-tuning LLM scaling factors appear to be very task dependent; however, there has been little published work examining fine-tuning focusing on the e-commerce domain. There has been some prior work investigating the use of LLMs on e-commerce tasks. ECOMGPT (Li et al . , 2023 ) looked at framing e-commerce tasks as instruction fine tuning, but doesn’t explore LoRA (Hu et al . , 2021 ) or how different tasks enhance or interfere with each other or the amount of data required for reasonable performance.
3. E-Commerce Datasets
There are a limited number of e-commerce datasets publicly available. Currently, there are few e-commerce benchmarks for evaluating LLMs on e-commerce tasks. We collected four datasets covering classification, sequence labeling, and product description generation, and review summarization in order to evaluate the performance of LLMs in the e-commerce domain.
Task | Train | Dev | Test |
ESCI Classification | 1,393,063 | - | 425,762 |
QueryNER: Query Segmentation | 7,841 | 871 | 933 |
Review Summarization | 25,203 | 3,114 | 3,166 |
Description Generation | 431,470 | - | 103,865 |
3.1. ESCI Multi-class Product Classification
The Shopping Queries ESCI dataset (Reddy et al . , 2023 ) contains search queries, released with the aim of fostering research in the area of semantic matching of queries and products. The dataset contains three tasks: Query-Product Ranking, Multi-Class Product Classification, Product Substitute Identification. We use the ESCI Multi-Class Product Classification task. The task is to classify a query and product pair as an exact match (E), a substitute (S), a complementary product (C) or Irrelevant (I). Because Query-Product Ranking and Product Substitute Identification involve more complexity and longer input strings, we do not include them for LLM evaluation in this work.
3.2. QueryNER
QueryNER (Palen-Michel et al . , 2024 ) is an e-commerce query segmentation dataset. The task in QueryNER is not to extract aspects, but rather to segment the user’s query into meaningful chunks. QueryNER uses a subset of the Amazon Shopping Queries Dataset (Reddy et al . , 2023 ) as the underlying data. QueryNER consists of an ontology of 17 types. The entity types are: core_product_type, product_name, product_number, modifier, creator, condition, unit of measurement (UoM), department, material, time, content, color, shape, quantity, occasion, origin, and price.
Because QueryNER uses BIO encoding ( B for begin, I for inside a span, O for outside a span), we linearized the data in order to have an input prompt and output string. The formatting of the linearized data is a series of (token, label) pairs.
3.3. Review Summarization
AMASUM (Bražinskas et al . , 2021 ) is a dataset for summarizing product reviews. The product reviews are in English and come from bestreviews.com, cnet.com, pmag.co.uk, runrepeat.com, which mainly consist of electronics and running shoes reviews. The dataset contains a list of product reviews and a summary with the “verdict” on the product and also lists of pros and cons. The original paper focuses on selecting useful reviews in order to summarize.
The goal of our evaluation of LLM performance is not to assess its review selection capability, so we select a small number of reviews to summarize. We select only 4 reviews to be used to generate the verdict summary. Since the dataset has a field with helpful votes where users voted that the review was helpful, we take the top four reviews with the most helpful votes as our selection process.
3.4. Product Description Generation
There appears to be a lack of standard benchmark datasets in English for product description generation despite a decent amount of prior work. Koto et al . ( 2022 ) stated they were not able to release the dataset due to copyright issues. Chan et al . ( 2019 ) and Zhang et al . ( 2019 ) collected data from Taobao 1 1 1 https://www.taobao.com/ . Zhang et al . ( 2019 ) also stated that there was no other standard dataset for product description generation. Wang et al . ( 2017 ) created their dataset from attribute values and descriptions from Amazon but only in the category ”Computers and Tablets”.
Without a clear prior benchmark for this task, we create a simple product description task from the ESCI Shopping Queries Dataset (Reddy et al . , 2023 ) . We assemble an input consisting of a product title, brand, color, and bullet points. The bullet points in the original dataset are aspect-value pair information about a product or short snippets about the product. The expected output is the product description. We filter out items where there is no title, no description, no bullet points, or items where the description is an exact match of the title or bullet points.
3.5. Task Alignment & Prompt Design
To enable instruction fine-tuning, each individual dataset was required to be aligned to a sequence to sequence style task in order to be used with an LLM. Review summarization and product description generation already were easily treated as sequence to sequence tasks. However it was less obvious how to best treat classification and sequence labeling tasks. We treated the classification label as text to be generated. For sequence labeling, the model was expected to output tuples of each token along with its label.
Prompts were designed to provide the model with enough context to accomplish its task. We created task specific templates for each of the tasks. Each prompt asked the model to act as an e-commerce expert to provide context. The prompt then gave a brief description of the task and along with the training example. Examples of prompts are shown in Appendix Appendix .
4. Experiments
4.1. baselines.
We chose to use the most common and competitive baselines for each e-commerce task. For classification tasks, ESCI Task 2 and QueryNER, we chose to use BERT (BERT-base) (Devlin et al . , 2019 ) as the baseline model with a learning rate of 3e-5 following Devlin et al . ( 2019 ) . The training of BERT followed the conventional formulation of the sequence classification problem for the ESCI task and token classification for the QueryNER task.
For generative tasks, review summarization, and product description generation, we chose to use T5 (T5-base) (Raffel et al . , 2020 ) as a baseline model. T5 and BERT are fine-tuned with default parameters released by Hugging face (Wolf et al . , 2020 ) . Because in-context learning has been shown to be effective (Dong et al . , 2024 ) , we use we include Mixtral 8 x 22b (Jiang et al . , 2024 ) as a theoretical state of the art zero-shot and few-shot baseline.
4.2. Mix Tasks Training
Mixing tasks (datasets) for fine-tuning large language models (LLMs) can enhance the model’s performance, generalization ability, and adaptability to various tasks, which closely mirrors the industrial application. The trained model was not required to accomplish one task but rather several domain-specific tasks such as query named entity recognition (NER), text summarization, description generation, and classification. Fine-tuning LLMs with mixed and diverse training datasets could help improve performance on each task.
4.3. Mix LoRA Merging
𝐖 1 𝑛 superscript subscript 𝑖 0 𝑛 1 subscript 𝐀 𝑖 superscript subscript 𝐁 𝑖 𝑇 \mathbf{W}^{\prime}=\mathbf{W}+\frac{1}{n}\sum_{i=0}^{n-1}\mathbf{A}_{i}% \mathbf{B}_{i}^{T} bold_W start_POSTSUPERSCRIPT ′ end_POSTSUPERSCRIPT = bold_W + divide start_ARG 1 end_ARG start_ARG italic_n end_ARG ∑ start_POSTSUBSCRIPT italic_i = 0 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT italic_n - 1 end_POSTSUPERSCRIPT bold_A start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT bold_B start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT start_POSTSUPERSCRIPT italic_T end_POSTSUPERSCRIPT . Mixing LoRA merging provided additional flexibility since additional tasks could be added later instead of retraining whenever a new task was added. Additionally, some types of tasks may benefit each other, while others may lower another tasks performance when merged. Mixing LoRA merge enabled more efficient experimentation with different combinations of tasks compared with directly training with the mixed data set.
4.4. Implementation Details
The foundation model used in all LoRA fine-tuning experiments was Llama-2-7b as its moderate size and considerable performance on various tasks (Touvron et al . , 2023 ) . The supervised fine-tune was conducted with 3 3 3 3 epochs for each dataset, as we do not see performance gain while we trained it longer. Since all four tasks for the LLM were formulated as a text generation task (see Section 3.5 ), we followed the most common hyperparameters to finetune the LLM under the text generation setup. Specifically, the model was loaded with 8 8 8 8 bit, the max length of input was set to 1500 1500 1500 1500 , and the precision of the parameter was set to be bf16 . During the supervised fine-tune, we adopted the LoRA (Hu et al . , 2021 ) to conduct efficient model training. The LoRA α 𝛼 \alpha italic_α was set to be 16 16 16 16 , with dropout rate be 0.05 0.05 0.05 0.05 . The initial learning rate was chosen to be 3 e − 5 3 𝑒 5 3e-5 3 italic_e - 5 with a cosine learning rate scheduler. To further improve the training efficiency, we adopted paged_adamw_8bit as optimizer. The LLM model training were conducted on two NVIDIA A100 80GB GPUs.
To optimize computational time, we randomly sampled subsets from each dataset for our experiments. Specifically, for the ESCI dataset, we used samples of 5k, 10k and 50k as the training set and sampled 1k from the original test set to serve as the test set. For the QueryNER dataset, we selected 1k, 5k, and 8k (full 7,841 data samples) samples for training and used the entire original test set as the test set. In the case of the description generation dataset, we randomly chose 1k, 5k, 10k, and 25k samples for training, with 10k samples from the original test set used as the test set. Finally, for the Review Summarization dataset, we sampled 1k, 5k, 10k, and 25k (the entire 25,203 dataset) as the training set and utilized the complete original test set as the test set. All experiments were conducted using the same aforementioned hyperparameters.
5.1. Metrics on different tasks
Since the datasets we explored contained both classification and generative tasks, we use the task-specific metric to evaluate the performance of the model performance on various datasets. For classification tasks, we use F1 score as evaluation metrics and report both micro-average and macro-average results, while for the generative task, we use the Rouge-1 F1 (Rouge-1) and Rouge-L F1 (Rouge-L) (Lin and Och, 2004 ) to evaluate the performance.
5.2. Evaluating LLM on Classification Tasks
To ensure accurate mapping of the generated text to the actual class label in the classification task, we implemented a simple but strict evaluation approach. For ESCI classification, the LLM output was required to match the corresponding label exactly. Any deviation from the exact label resulted in the classification being considered incorrect. For QueryNER, labels were extracted using a regular expression expecting a list of tuples of tokens with BIO tags. If the LLM output deviated from the expected output, the labels were considered all O s (no entities identified). Because we noticed the model sometimes being inconsistent with the output format, the regular expression also handles comma separated output without parentheses. In the case of further formatting issues or when the model does not predict labels for all tokens, it is assumed the model failed to generate a valid label sequence and that particular example is assigned all O s. Palen-Michel et al . ( 2021 ) highlighted issues with NER scoring. For scoring procedure clarity, once the model output has been extracted from the linearized form, we use seqeval (Nakayama, 2018 ) for evaluation using the setting which is equivalent with conlleval.
5.3. SFT LLMs vs Baseline PLMs
We performed SFT training of Llama2-7b on each dataset with different portions of the data to explore the impact of training data size for fine-tuning an LLM on each task.
5.3.1. Classification tasks
ESCI Classification | QueryNER | ||||
Micro F1 | Macro F1 | Micro F1 | Macro F1 | ||
BERT @5k | 0.348 | 0.181 | @1k | 0.539 | 0.390 |
LLM SFT @5k | 0.355 | 0.244 | @1k | 0.280 | 0.156 |
BERT @10k | 0.628 | 0.294 | @5k | 0.580 | 0.508 |
LLM SFT @10k | 0.397 | 0.213 | @5k | 0.553 | 0.398 |
BERT @50k | 0.629 | 0.368 | @8k | 0.603 | 0.569 |
LLM SFT @50k | 0.628 | 0.294 | @8k | 0.626 | 0.579 |
Mixtral 0-shot | 0.571 | 0.199 | 0.145 | 0.063 | |
Mixtral 3-shot | 0.537 | 0.009 | 0.484 | 0.336 |
Table 2 shows the performance comparison on classification tasks on ESCI Classification and QueryNER dataset among Llama2-7b Supervised fine-tuning, BERT model fine-tuning and in context learning using Mixtral 8 x 22b in zero and few-shot setup. Note that, instead of generating the distribution like BERT, the task for the LLM is to generate the classification result in text. As the number of the training samples increased the performance of the model generally increased. However, there was a clear performance boost of LLM as the number of training samples increased (from 10k to 50k on ESCI task 2 dataset and from 1k to 5k on Query NER dataset). In general, the LLM and BERT performed comparable in these classification tasks when given sufficient training data.
In domain-specific tasks such as ESCI classification and Query NER, the application of in-context learning with very large language models like the Mixtral 8x22b often does not meet the performance benchmarks achieved through fine-tuning. Despite the introduction of extensive context, these models frequently struggle to deliver the level of accuracy required for industrial applications. This observation underscores a critical limitation: while LLMs are versatile and powerful, they may not be inherently optimized for tasks that demand high precision within a specialized domain.
In contrast, fine-tuning enables models to be specifically tailored to the intricacies of the task at hand, allowing for a deeper understanding of domain-specific patterns and nuances. As a result, training smaller, task-specific models such as BERT, particularly those employing a softmax classification layer, often emerges as a more effective strategy. These models not only demonstrate superior performance but also offer advantages in computational efficiency, making them more suitable for deployment in resource-constrained industrial environments where both accuracy and efficiency are paramount.
5.3.2. Generation task
Review Summarization | Desc. Generation | |||
Rouge-1 | Rouge-L | Rouge-1 | Rouge-L | |
T5 @1k | 0.155 | 0.137 | 0.239 | 0.216 |
LLM SFT @1k | 0.158 | 0.147 | 0.262 | 0.244 |
T5 @5k | 0.162 | 0.147 | 0.241 | 0.223 |
LLM SFT @5k | 0.182 | 0.161 | 0.249 | 0.232 |
T5 @10k | 0.163 | 0.150 | 0.238 | 0.221 |
LLM SFT @10k | 0.186 | 0.162 | 0.258 | 0.241 |
T5 @25k | 0.169 | 0.158 | 0.232 | 0.215 |
LLM SFT @25k | 0.187 | 0.165 | 0.237 | 0.222 |
Mixtral 0-shot | 0.099 | 0.090 | 0.248 | 0.229 |
Mixtral 3-shot | 0.144 | 0.131 | 0.274 | 0.254 |
Table 3 shows the performance comparison of text generation tasks on the review summarization and description generation datasets. Similar to the classification tasks, there was a significant increase in model performance as more data samples were used for training. Notably, the LLM consistently outperformed the conventional T5 model across both datasets. This superior performance can be attributed to the LLM’s larger model capacity and enhanced quality of pre-training.
Fine-tuned models (Llama2-7b and T5) outperformed the zero-shot capabilities of the much larger Mixtral 8x22B model in review summarization, while for description generation, the performance is comparable. Despite the Mixtral model’s strong standing on LLM leaderboards like (Fourrier et al . , 2024 ) , which suggests competitive summarization abilities, the observed performance gap between zero-shot and few-shot scenarios highlights a key limitation: without in-context guidance, the model struggles to achieve sufficient capability on domain-specific tasks (Review Summarization). However, when in-context information is provided, the model demonstrated significantly improved outcomes. In review summarization, to achieve even higher levels of performance, task-specific training becomes crucial. Notably, even with smaller model architectures, fine-tuning can yield superior results (using Llama2-7b).
In contrast, the description generation task is more aligned with general-purpose text generation, where the model’s ability to understand and leverage general knowledge is the primary factor in determining performance. Consequently, in this task, larger models like Mixtral, equipped with in-context guidance, could achieve top-tier performance, even surpassing fine-tuned smaller models.
QueryNER | Review Summ. | Desc. Generation | ||||
Micro F1 | Marco F1 | Rouge-1 | Rouge-L | Rouge-1 | Rouge-L | |
QueryNER @ 5k | 0.553 | 0.398 | - | - | - | - |
+ Summ. LoRA @ 5k | 0.002 | 0.232 | 0.192 | 0.164 | - | - |
+ Desc. Generation @ 5k | 0.018 | 0.344 | - | - | 0.251 | 0.233 |
ESCI | Review Summ. | Desc. Generation | ||||
ESCI @ 5k | 0.355 | 0.244 | - | - | - | - |
+ Summ. LoRA @ 5k | 0.145 | 0.174 | 0.184 | 0.156 | - | - |
+ Desc. Generation LoRA @ 5k | 0.137 | 0.299 | - | - | 0.239 | 0.221 |
5.4. LoRA Merge
We experimented with merging different pairs of LoRA weights for each pair of tasks. For this experiment, we used the LoRA weights from the 5k training set size. To merge the LoRA weights we took the average of the two. The results of merging LoRA weights, shown in Table 4 , demonstrated that when weights trained on a task requiring a more rigid structure in the output like ESCI classification or QueryNER, the performance on those tasks suffers. However, the performance of description generation and review summarization remained comparable with the performance from independent training with the same number of examples.
Upon reviewing model output, we found that at least a portion of this degradation in performance on the tasks requiring a more strict output format was attributable to the output formatting requirements. However, some of this apparent degradation may not truly be quite as bad as it appears.
In Section C of the appendix, we show an example of model output for the QueryNER task was shown where the model did in fact output BIO labels some of which are correct labels, but with formatting unrecognized by the scoring script.
IMAGES
VIDEO
COMMENTS
Electronic Commerce Research is a peer-reviewed journal that publishes theoretical and empirical research on e-commerce. It covers topics such as AI, big data, green supply chain, consumer happiness, and cybersecurity.
ECRA is a peer-reviewed journal that publishes papers on various topics related to e-commerce, such as technology, business, policy, and analytics. It aims to create and disseminate enduring knowledge for the fast-changing e-commerce environment and invites proposals for special issues on new topics.
Search all Electronic Commerce Research articles Volume 24 March - June 2024 Mar - Jun 2024. Issue 2 June 2024. Special Issue : Online Grocery Shopping. Issue 1 March 2024. SI: Electronic Commerce in Finance. Volume 23 March - December 2023 Mar - Dec 2023. Issue 4 December 2023; Issue 3 September 2023.
Browse the latest articles on e-commerce research from various perspectives, such as marketing, technology, psychology, and management. Find out the aims, scope, and editorial board of the journal, as well as the submission guidelines and updates.
The International Journal of Electronic Commerce is the leading refereed quarterly devoted to advancing the understanding and practice of electronic commerce. It serves the needs of researchers as well as practitioners and executives involved in electronic commerce. The Journal aims to offer an integrated view of the field by presenting approaches of multiple disciplines.
JECR is a peer-reviewed journal that publishes research on e-commerce topics. The current issue features articles on grammar and syntax, customer satisfaction, social media influencers, and a special issue on artificial general intelligence in e-commerce.
A peer-reviewed journal published by Springer Science+Business Media that covers all aspects of electronic commerce. The journal has an impact factor of 3.7 and is indexed in various databases and platforms.
Electronic Commerce Research and Applications aims to create and disseminate enduring knowledge for the fast-changing e-commerce environment. A major dilemma in e-commerce research is how to achieve a balance between the currency and the life span of knowledge. Electronic Commerce Research and Applications will contribute to the establishment ...
A peer-reviewed journal that publishes research on e-commerce technologies, theory, applications, and policy. Covers topics such as agent-based commerce, big data analytics, social media, digital economy, e-government, and more.
A collection of research articles on various topics in e-commerce, such as platform advertising, service quality, personalization, live streaming, co-creation, recommendation systems, and more. The articles are published in the journal Electronic Commerce Research and Applications, a peer-reviewed scholarly platform by Elsevier.
2021 marks the 20th anniversary of the founding of Electronic Commerce Research ( ECR ). The journal has changed substantially over its life, reflecting the wider changes in the tools and commercial focus of electronic commerce. ECR 's early focus was telecommunications and electronic commerce. After reorganization and new editorship in 2014 ...
A peer-reviewed journal that publishes research on e-commerce topics such as information systems, marketing, artificial intelligence, and computer networks. Learn about the aims, scope, impact, and publishing options of this journal.
Information transmit strategy of e-commerce platform with financially constrained supplier. Zhaobo Chen. Research Center for Innovation and Development of Equipment Manufacturing Industry, Taiyuan University of Science and Technology, Taiyuan 030024, China
Journal of Theoretical and Applied Electronic Commerce Research published since 2006, is an international, peer-reviewed, scientific journal owned by the Faculty of Engineering of the Universidad de Talca, and MDPI provides publishing services for the journal since Volume 16, Issue 3 (2021).. Open Access — free for readers, with article processing charges (APC) paid by authors or their ...
Economics and Electronic Commerce. Google is a global leader in electronic commerce. Not surprisingly, considerable attention is devoted to research in this area. Topics include 1) auction design, 2) advertising effectiveness, 3) statistical methods, 4) forecasting and prediction, 5) survey research, 6) policy analysis and a host of other topics.
Abstract. 2021 marks the 20th anniversary of the founding of Electronic Commerce Research ( ECR ). The journal has changed substantially over its life, reflecting the wider changes in the tools and commercial focus of electronic commerce. ECR 's early focus was telecommunications and electronic commerce. After reorganization and new ...
Electronic Commerce Research is a hybrid journal that publishes theoretical and empirical research on all facets of electronic commerce and its implications. The journal covers topics such as Internet services, intelligent agents, global impact, fraud reduction, security, payment systems, marketing, legal issues, and more.
All articles published in Journal of Theoretical and Applied Electronic Commerce Research (ISSN 0718-1876) are published in full open access.An article processing charge (APC) of CHF 1000 (Swiss Francs) applies to papers accepted after peer review. This article processing charge is to cover the costs of peer review, copyediting, typesetting, long-term archiving, and journal management.
Author Correction: Online Advertising and Real Estate sales: evidence from the Housing Market. Xiuzhi Zhang. Ying Zhang. Zhijie Lin. Author Correction Open access 06 March 2023 Pages: 631 - 631. Volume 23, issue 1 articles listing for Electronic Commerce Research.
This is a complete update of the best-selling undergraduate textbook on Electronic Commerce (EC). New to this 4th Edition is the addition of material on Social Commerce (two chapters); a new tutorial on the major EC support technologies, including cloud computing, RFID, and EDI; ten new learning outcomes; and video exercises added to most chapters. Wherever appropriate, material on Social ...
In recent years, the development of e-commerce platforms has been gaining momentum, which has made the mining and marketing of e-commerce data more and more critical. Major e-commerce platforms and companies are starting to analyse basic information about their customers, extracting useful information from it, accurately screening customer needs and finally developing targeted marketing ...
JOURNAL OF ELECTRONIC COMMERCE RESEARCH. College of Business - California State University Long Beach . 1250 Bellflower Blvd, Long Beach, CA 90840. Hosted & Printed by College of Commerce, National Chengchi University. 64, Sec. 2, ZhiNan Rd., Taipei, Taiwan 11605.
ELECTRONIC COMMERCE IN THE AGE OF DIGITALIZATION. May 2020. Black Sea Economic Studies. DOI: 10.32843/bses.53-8. Authors: Olena Vynogradova. Natalia Yevtushenko. Iryna Krjuchok. To read the full ...
Our contributions are as follows: 1) Organizing and formatting four e-commerce datasets for the evaluation of large language models (LLMs). 2) Conducting comprehensive experiments to compare fine-tuning a large language model with conventional industrial baselines, such as BERT and T5, using varying amounts of data for e-commerce tasks; additionally, we examine the effectiveness of in-context ...
2021 marks the 20th anniversary of the founding of Electronic Commerce Research (ECR). The journal has changed substantially over its life, reflecting the wider changes in the tools and commercial focus of electronic commerce. ECR's early focus was telecommunications and electronic commerce. After reorganization and new editorship in 2014, that focus expanded to embrace emerging tools ...
Our research attempts to provide a theoretical basis for China's consumer goods e-commerce market entry barriers. Practical implications Studying market entry barriers can help firms better adapt to market changes, optimize their operational modes according to market demands and norms, and enhance their competitiveness.
The global cross-border e-commerce logistics market size was estimated at USD 97.85 billion in 2023 and is projected to grow at a CAGR of 25.4% from 2024 to 2030. The market growth can be attributed to the rapid expansion of online retail, cross-border trade, and advancements in logistics technology. ... Grand View Research has segmented the ...
After five years of negotiations, 80 member countries of the World Trade Organization (WTO) reached an historic result on cross-border e-commerce, agreeing on a draft of common guidelines and ...