Generating Product Descriptions from User Reviews
Open accessThe DOI matches its Crossref record; 72 of the 73 references it cites were found in a registry or an index.
Abstract
Product descriptions play an important role in the e-commerce ecosystem, conveying to buyers information about a merchandise they may purchase. Yet, on leading e-commerce websites, with high volumes of new items offered for sale every day, product descriptions are often lacking or missing altogether. Moreover, many descriptions include information that holds little value and sometimes even disrupts buyers, in an attempt to draw attention and purchases. In this work, we suggest to mitigate these issues by generating short crowd-based product descriptions from user reviews . We apply an extractive approach, where review sentences are used in their original form to compose the product description. At the core of our method is a supervised approach to identify candidate review sentences suitable to be used as part of a description. Our analysis, based on data from both the Fashion and Motors domains, reveals the top reasons for review sentences being unsuitable for the product's description and these are used, in turn, as part of a deep multi-task learning architecture. We then diversify the set of candidates by removing redundancies and, at the final step, select the top candidates to be included in the description. We compare different methods for each step and also conduct an end-to-end evaluation, based on rating from professional annotators, showing the generated descriptions are of high quality.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“Prior work towards general users investigated extracting experiences [21, 24], tips [11, 41] or snippets suitable for product descriptions [7, 25], yet it remains unknown how reviews can help with the unique interests of BLV users, especially providing descriptive information on product appearance.”
From Revamp: Enhancing Accessible Information Seeking Experience of Online Shopping for Blind or Low Vision Users (2021) · cites this paper as [25]
In the citing paper
“if we follow prior work on filtering out subjective comments [25], we will lose the vivid and descriptive details that are crucial to appearance understanding.”
From Revamp: Enhancing Accessible Information Seeking Experience of Online Shopping for Blind or Low Vision Users (2021) · cites this paper as [25]
Citations
Cites
Hochreiter, S., & Schmidhuber, J. (1997) · Neural Computation · doi:10.1162/neco.1997.9.8.1735 · Cited by 102,318
XGBoost: A Scalable Tree Boosting System
Chen, T., & Guestrin, C. (2016) · Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining · doi:10.1145/2939672.2939785 · Cited by 54,569
A Coefficient of Agreement for Nominal Scales
Cohen, J. (1960) · Educational and Psychological Measurement · doi:10.1177/001316446002000104 · Cited by 42,398 · cited 1×
“At the end of the process, the inter-annotator agreement for the task of good versus bad labeling, measured by Cohen’s kappa [9], was 0 .”
MethodGirshick, R. (2015) · 2015 IEEE International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2015.169 · Cited by 28,105
Efficient Estimation of Word Representations in Vector Space
Mikolov, T., Chen, K., Corrado, G., & Dean, J. (2013) · arXiv · doi:10.48550/arxiv.1301.3781 · Cited by 26,568 · cited 1×
“A recurrent neural network based on long short-term memory (LSTM) [25] architecture, with pre-trained word2vec embeddings using continuous bag of words (CBOW) with negative sampling [49].”
Background(1967) · Cited by 23,719
Weighted kappa: nominal scale agreement provision for scaled disagreement or partial credit
Cohen, J. (1968) · Psychological Bulletin · doi:10.1037/h0026256 · Cited by 8,736 · cited 1×
“The inter-annotator agreement for the main quality evaluation question, measured using weighted Kappa [10], was 0 .”
MethodEffective Approaches to Attention-based Neural Machine Translation
Luong, T., Pham, H., & Manning, C. D. (2015) · Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing · doi:10.18653/v1/d15-1166 · Cited by 8,382 · cited 1×
“The effectiveness of the attention mechanism, however, typically lies in longer sentences [44].”
BackgroundMining and summarizing customer reviews
Hu, M., & Liu, B. (2004) · Proceedings of the tenth ACM SIGKDD international conference on Knowledge discovery and data mining · doi:10.1145/1014052.1014073 · Cited by 7,793 · cited 1×
“, [26, 45, 71]), however such summaries do not necessarily contain descriptions of the product.”
BackgroundOpinion Mining and Sentiment Analysis
Pang, B., & Lee, L. (2008) · now publishers, Inc. eBooks · doi:10.1561/9781601981516 · Cited by 6,720
Research works printed but not linked: 7
- 10.5555/1622487.1622501 · doi:10.5555/1622487.1622501
- Gagandeep Kaur and Gagandeep Kaur . 2016 . Mobile applications are major players in the world of e-commerce . International Journal of Advanced Research in IT and Engineering 5 , 2(2016), 13 - 21 . Gagandeep Kaur and Gagandeep Kaur. 2016. Mobile applications are major players in the world of e-commerce. International Journal of Advanced Research in IT and Engineering 5, 2(2016), 13-21.
- Zehra Kavasoglu and Sule Gündüz Ögüdücü . 2013 . Personalized summarization of customer reviews based on user's browsing history . IADIS International Journal on Computer Science & Information Systems 8 , 2(2013), 147 - 158 . Zehra Kavasoglu and Sule Gündüz Ögüdücü. 2013. Personalized summarization of customer reviews based on user's browsing history. IADIS International Journal on Computer Science & Information Systems 8, 2(2013), 147-158.
- Diederik P. Kingma and Jimmy Ba . 2014 . Adam : A method for stochastic optimization. arXiv preprint abs/1412.6980 (2014). Diederik P. Kingma and Jimmy Ba. 2014. Adam: A method for stochastic optimization. arXiv preprint abs/1412.6980 (2014). · arXiv
- 10.5555/2886521.2886636 · doi:10.5555/2886521.2886636
- Ajinkya More . 2016. Attribute extraction from product titles in eCommerce. arXiv preprint abs/1608.04670 ( 2016 ). Ajinkya More. 2016. Attribute extraction from product titles in eCommerce. arXiv preprint abs/1608.04670 (2016). · arXiv
- Tijmen Tieleman and Geoffrey Hinton . 2012. Lecture 6.5-rmsprop : Divide the gradient by a running average of its recent magnitude. COURSERA: Neural networks for machine learning 4, 2 ( 2012 ), 26-31. Tijmen Tieleman and Geoffrey Hinton. 2012. Lecture 6.5-rmsprop: Divide the gradient by a running average of its recent magnitude. COURSERA: Neural networks for machine learning 4, 2 (2012), 26-31.
Cited by
Artificial Intelligence in Business-to-Customer Fashion Retail: A Literature Review
Goti, A., Lomas, L. Q., Almeida, A., de la Puerta, J. G., & López–de–Ipiña, D. (2023) · Mathematics · doi:10.3390/math11132943 · Cited by 92
Online User Review Analysis for Product Evaluation and Improvement
Yang, C., Wu, L., Tan, K., Yu, C., Zhou, Y., Tao, Y., & Song, Y. (2021) · Journal of theoretical and applied electronic commerce research · doi:10.3390/jtaer16050090 · Cited by 44
“[22] extract product descriptions from user reviews using deep multi-task learning.”
BackgroundWang, R., Chen, Z., Zhang, M. R., Li, Z., Liu, Z., Dang, Z., Yu, C., & Chen, X. '. (2021) · Proceedings of the 2021 CHI Conference on Human Factors in Computing Systems · doi:10.1145/3411764.3445547 · Cited by 42
“Prior work towards general users investigated extracting experiences [21, 24], tips [11, 41] or snippets suitable for product descriptions [7, 25], yet it remains”
BackgroundMethodTAERT: Triple-Attentional Explainable Recommendation with Temporal Convolutional Network
Guo, S., Wang, Y., Yuan, H., Huang, Z., Chen, J., & Wang, X. (2021) · Information Sciences · doi:10.1016/j.ins.2021.03.034 · Cited by 37
Category Recognition in E-Commerce using Sequence-to-Sequence Hierarchical Classification
Hasson, I., Novgorodov, S., Fuchs, G., & Acriche, Y. (2021) · ACM International Conference on Web Search and Data Mining (WSDM) · doi:10.1145/3437963.3441717 · Cited by 29
“, image, title, attributes and description [12]), and its effect on the listing’s likelihood to sell.”
BackgroundSelf-citationReqGen: Keywords-Driven Software Requirements Generation
Zhao, Z., Zhang, L., Lian, X., Gao, X., Lv, H., & Shi, L. (2023) · Mathematics · doi:10.3390/math11020332 · Cited by 18
Persuade to Click: Context-aware Persuasion Model for Online Textual Advertisement
Yuan, Y., Xu, F., Cao, H., Zhang, G., Hui, P., Li, Y., & Jin, D. (2021) · IEEE Transactions on Knowledge and Data Engineering · doi:10.1109/tkde.2021.3110724 · Cited by 19
“There are also a wide range of studies on automated text generation, including health-related advertisement derived from web pages [66], product descriptions [1], [2] and review response [3] on e-commerce websites.”
BackgroundGenerating Tips from Product Reviews
Hirsch, S., Novgorodov, S., Guy, I., & Nus, A. (2021) · ACM International Conference on Web Search and Data Mining (WSDM) · doi:10.1145/3437963.3441755 · Cited by 15
“scription generation from reviews [30] and can be similarly applied in our case.”
BackgroundMethodInfluentialSelf-citationProduct Bundle Identification using Semi-Supervised Learning
Tzaban, H., Guy, I., Greenstein‐Messica, A., Dagan, A., Rokach, L., & Shapira, B. (2020) · International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR) · doi:10.1145/3397271.3401128 · Cited by 19
“We did not examine the use of product descriptions [12, 39] for the bundle identification task, and leave this direction for future work.”
BackgroundSelf-citationLearning to Generate Personalized Product Descriptions
Elad, G., Guy, I., Novgorodov, S., Kimelfeld, B., & Radinsky, K. (2019) · Proceedings of the 28th ACM International Conference on Information and Knowledge Management · doi:10.1145/3357384.3357984 · Cited by 13
“[44] proposed a method for generating non-personalized product descriptions from user reviews.”
MethodSelf-citationNovgorodov, S., Guy, I., Elad, G., & Radinsky, K. (2020) · ACM Transactions on Internet Technology · doi:10.1145/3418202 · Cited by 19
“Our previous work [59] addresses the same problem of description generation from reviews, however, the current article provides additional analysis, extended set of experiments, and initial A/B testing results.”
Self-citationTowards Extracting Highlights From Recorded Live Videos: An Implicit Crowdsourcing Approach
Jiang, R., Qu, C., Wang, J., Wang, C., & Zheng, Y. (2020) · IEEE International Conference on Data Engineering (ICDE) · doi:10.1109/icde48307.2020.00176 · Cited by 14
“However, prior work is mainly focused on explicit crowdsourcing [20]–[22] or non-video-tasks [23]–[27].”
BackgroundGenerating Product Description with Generative Pre-trained Transformer 2
Nguyen, M.-T., Nguyen, P.-T., Nguyen, V.-V., & Nguyen, Q.-M. (2021) · International Conference on Innovative Technologies in Intelligent Systems and Industrial Applications (CITISIA) · doi:10.1109/citisia53721.2021.9719940 · Cited by 6
Hao, S., Guo, B., Wang, H., Liang, Y., Yao, L., Wang, Q., & Yu, Z. (2021) · ACM Transactions on Knowledge Discovery from Data · doi:10.1145/3446982 · Cited by 6
“[18] proposed a rule-based solution to select appropriate sentences from reviews and applied the LexRank algorithm to produce the final product description.”
BackgroundMethodDescription Generation for Points of Interest
Zhou, M. R., Zhou, J., Fu, Y., Ren, Z., Wang, X., & Xiong, H. (2021) · IEEE International Conference on Data Engineering (ICDE) · doi:10.1109/icde51399.2021.00226 · Cited by 6
“Reference to the research conducted by Novgorodov et al. [28], some high-quality reviews can be used directly as ground-truth for description generation.”
MethodDeep or simple models for semantic tagging?
Li, J., Li, Y., Wang, X., & Tan, W.-C. (2020) · Proceedings of the VLDB Endowment · doi:10.14778/3407790.3407844 · Cited by 6
“Some recent studies [37, 36, 50, 17, 52, 39, 35, 20, 47, 53] made comparisons between deep models and simple models (i.”
BackgroundResultLi, Q., Li, P., Li, X., Ren, Z., Chen, Z., & de Rijke, M. (2021) · ACM International Conference on Web Search and Data Mining (WSDM) · doi:10.1145/3437963.3441804 · Cited by 6
“Many studies have focused on mining valuable information from reviews and shown promising results in various tasks, such as opinion summarization [1, 2, 5, 8, 24, 25, 40] and item description generation [13, 35].”
BackgroundInterpretable Aspect-Aware Capsule Network for Peer Review Based Citation Count Prediction
Li, S., Li, Y., Zhao, W. X., Ding, B., & Wen, J.-R. (2021) · ACM Transactions on Information Systems · doi:10.1145/3466640 · Cited by 7
“However, research works on a similar task of predicting the quality of products using review text are relatively rich [8, 17, 28, 35, 36, 39, 47, 60, 63, 68, 69, 79, 82–86].”
BackgroundRodriguez-Diaz, C. A., Jimenez, S., Bejarano, D., Chávez, J. A. B., & Gelbukh, A. F. (2023) · Information Sciences · doi:10.1016/j.ins.2023.118953 · Cited by 4
Cross-Cultural Transfer Learning for Text Classification
Ringel, D., Lavee, G., Guy, I., & Radinsky, K. (2019) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/v1/d19-1400 · Cited by 4
“We consider the description texts to be examples of formal writing and the reviews to be examples of informal writing (Novgorodov et al., 2019).”
BackgroundSelf-citationAnggraeni, R. M., & Ismi, D. P. (2022) · Science in Information Technology Letters · doi:10.31763/sitech.v3i2.1225 · Cited by 3
Applications of Text Generation in Digital Marketing: a review
Loukili, S., Fennan, A., & Elaachak, L. (2023) · International Conference on Networking, Information Systems & Security (NISS) · doi:10.1145/3607720.3608451 · Cited by 8
Recurrent Point Processes for Dynamic Review ModelsPreprint
Cvejoski, K., Sánchez, R. J., Georgiev, B., Schuecker, J., Bauckhage, C., & Ojeda, C. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1912.04132 · Cited by 3
“The reviews research have sought to characterize usefulness and generation of reviews (Fan et al. 2019; Novgorodov et al. 2019) and provide better representations for rating prediction (Esmaeili et al.”
BackgroundThe Impact of Toxic Replies on Twitter Conversations
Salehabadi, N. (2019) · Cited by 3
Altammami, A., Dimitrova, V. G., & Pournaras, E. (2024) · International Conference on Information Technology for Social Good (GoodIT) · doi:10.1145/3677525.3678695 · Cited by 3
“In online shopping, people often struggle with processing extensive data [27, 48], primarily conveyed through textual product descriptions [34].”
Cvejoski, K., Sánchez, R. J., Georgiev, B., Bauckhage, C., & Ojeda, C. (2020) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn48605.2020.9206768 · Cited by 5
“They also provide characterization of review usefulness [20], use reviews for product description [21], and provide better representations for rating prediction [22].”
BackgroundLearning to Generate Tips from Song Reviews
Zang, J., Gao, C., Chen, Y., Xu, R., Zhou, L., & Wang, X. (2023) · Neural Networks · doi:10.1016/j.neunet.2023.01.049 · Cited by 2
Inventory Reduction via Maximal Coverage in E-Commerce
Gershtein, S., Milo, T., & Novgorodov, S. (2020) · Movebank · doi:10.5441/002/edbt.2020.63 · Cited by 4
“Some extensively studied problems are product classification [9, 33], product ranking in search results [18] and automatic product content generation [10, 24].”
BackgroundSelf-citationFelix, B., Gunawan, A. A. S., & Suhartono, D. (2024) · International Conference on Information Management and Technology (ICIMTech) · doi:10.1109/icimtech63123.2024.10780443 · Cited by 2
Leveraging AI and customer reviews to evaluate technology used by people with disabilities
Allen, T. G., Horwitz, A., & Sprigle, S. H. (2025) · Disability and Rehabilitation Assistive Technology · doi:10.1080/17483107.2025.2465603 · Cited by 1
Harrag, F., Touameur, O., & Behilil, H. (2025) · Performance Measurement and Metrics · doi:10.1108/pmm-12-2024-0058 · Cited by 1
Extracting Salient Facts from Company Reviews with Scarce Labels
Li, J., Bhutani, N., Whedon, A., Huang, C.-Y., Junior, E. R. H., & Suhara, Y. (2022) · AAAI Workshop on Artificial Intelligence with Biased or Scarce Data (AIBSD) · doi:10.3390/cmsf2022003009 · Cited by 1
“Word vector models [8,13,14,17,27,28],such as convolutional neural networks [29] and long short-term memory [30], represent a text as a matrix of word embeddings and can thereby process unseen words through their embeddings.”
BackgroundReducE-Comm: Effective Inventory Reduction System for E-Commerce
Gershtein, S., Milo, T., & Novgorodov, S. (2019) · ACM International Conference on Information and Knowledge Management (CIKM) · doi:10.1145/3357384.3357861
“Some of the extensively studied problems are product classification (e.g. [8]), ranking of products in search results (e.g. [5]), products recommendations (e.g [7]) and generation of product data (e.g. descriptions [6]).”
BackgroundSelf-citationDistilling Salient Reviews with Zero Labels
Huang, C.-Y., Li, J., Bhutani, N., Whedon, A., Junior, E. R. H., & Suhara, Y. (2022) · Fact Extraction and VERification Workshop (FEVER) · doi:10.18653/v1/2022.fever-1.3
“Most of existing extraction models (Novgorodov et al., 2019; Negi and Buitelaar, 2015; Guy et al., 2017a; Wang et al., 2019; Chen et al., 2014; Hua et al., 2019; Zhang et al., 2019; Gao et al., 2018; Li et al., 2019; Evensen et al., 2019) are supervised.”
BackgroundMethodLot or Not: Identifying Multi-Quantity Offerings in E-Commerce
Lavee, G., & Guy, I. (2022) · Workshop on e-Commerce and NLP (ECNLP) · doi:10.18653/v1/2022.ecnlp-1.29
“Another reason is broad applicability: while many offerings are incomplete to some degree, with lacking or altogether-missing attributes (Ghani et al., 2006), descriptions (Novgorodov et al., 2019) and images (Goswami et al., 2012), the vast majority of offerings contain a valid title.”
BackgroundSelf-citationAcademic Article Classification Algorithm Based on Pre-trained Model and Keyword Extraction
Zhou, Z., Zheng, D., Qiu, Z., Lin, R., Wu, Z., & Yuan, C. (2022) · Communications in computer and information science · doi:10.1007/978-981-19-4549-6_12
Zamani, B. (2023) · Emerald Group Publishing Limited eBooks · doi:10.1108/978-1-80382-233-420231009 · Cited by 1
Generating Product Insights from Community Q&A
Kuchy, L., Levy, R., Mejer, A., Segev, N., Agmon, S., & Farber, M. (2023) · ACM International Conference on Information and Knowledge Management (CIKM) · doi:10.1145/3583780.3615480
GERAÇÃO DE DESCRIÇÕES DE PRODUTOS A PARTIR DE AVALIAÇÕES DE USUÁRIOS USANDO UM LLM
GUTIERREZ, B. F. M. (2024) · doi:10.17771/pucrio.acad.66911
A Unified Platform for Summarizing E-Commerce Product Reviews Using Generative AI Tool
Shobhika, Shivam, Dubey, G., & Sharma, A. (2025) · International Conference on Innovations and Emerging Technologies in AI & Communication Systems (IETACS) · doi:10.1109/ietacs68750.2025.11385597
“Recent work by Novgorodov et al. (2019) emphasizes how large e-commerce sites often lack or have low-quality product descriptions and offers a new extractive method to combine the crowd-sourced product descriptions from user reviews.”
Ho, G. T. S., Wang, Y., Tang, V., Manviel, M. F., Elle, W. H., & Li, C. H. J. (2026) · International Journal of Engineering Business Management · doi:10.1177/18479790261462275
Metadata
- Authors
- Novgorodov, S., Guy, I., Elad, G., & Radinsky, K.
- Affiliations
- eBay Research, Israel
- Technion – Israel Institute of Technology (IL)
- Venue
- The World Wide Web Conference
- Publisher
- ACM
- Published
- 2019-05-13
- Pages
- 1354-1364
- OpenAlex
- W2912653541
- Cited by
- 41
- References
- 73
- Citations by year
- 2026: 1 · 2025: 3 · 2024: 3 · 2023: 7 · 2022: 6 · 2021: 11 · 2020: 5 · 2019: 5
- FWCI
- 2.68
- Type
- conference paper
- Language
- en
- Open access
- hybrid
- doi.org
- Licence
- Locations
- International World Wide Web Conference (WWW) (publishedVersion, cc-by)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Topic Modeling
- Physical Sciences › Computer Science › Artificial Intelligence › Sentiment Analysis and Opinion Mining
- Physical Sciences › Computer Science › Artificial Intelligence › Advanced Text Analysis Techniques
- Keywords
- product description generation, online reviews, crowdsourcing, extractive summarization, multi-task learning, redundancy elimination, e-commerce, fashion products
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.