Knowledge Management in e-Commerce

April 16: 10AM- 6PM CEST

Will be held in conjunction with the Web Conference 2021

All presentations will be virtual

The global e-Commerce market size is valued at USD 9.09 trillion with an annual growth rate of 14.7%. The 2020 pandemic dramatically changed people's lifestyles. E-Commerce will further accelerate its growth and penetration into people's daily lives. E-Commerce websites and apps are among the top visits of everyone's daily routine. Customers want E-Commerce websites and apps as their personal assistant that finds the exact products they are searching for, provides recommendations when they are not sure which products to buy, and answers questions about product details. Extracting structural knowledge about e-Commerce products from their text descriptions, images, reviews, customer interaction logs is the key for building delightful shopping experience for search, recommendation, advertising, and product QA. Many challenges in building a product knowledge base can benefit from the learnings of building a semantic web. On the other hand, the unique data in e-commerce can spike new research directions in the web conference community. The KMEcommerce workshop aims to bring together researchers from both academia and industry labs to exchange notes and get a pulse for the state of art of improving e-commerce customer experience with product knowledge mining and management.



  • Bing Yin
  • Luna Xin Dong
  • Haixun Wang
  • Chao Zhang
  • Sreyashi Nag
  • Diego Marcheggiani

Program Committee

  • Diego Marcheggiani
  • Yunwen Xu
  • Heran Lin
    Tsinghua University
  • Rahul Goutam
  • Jun Ma
  • Nick Blumm
  • Tianyu Cao
  • Vijai Mohan
  • Youna Hu
  • Tracy Holloway King
  • Jiufeng Zhou

Key Dates

• 12/21/2020: Submissions open
• 2/28/2021: Workshop papers due
• 3/21/2021: Notification of accepted papers
• 4/16/2021: Workshop Day


CET 10:00am/PDT 1:00am/CDT 3:00am/Beijing 4:00pm - Opening
CET 10:15am/PDT 1:15am/CDT 3:15am/Beijing 4:15pm - Contributed Talks Session 1

CET 11:30am/PDT 2:30am/CDT 4:30am/Beijing 5:30pm - Invited talk 'AliCoCo: Alibaba E-commerce Cognitive Concept Net' by Xusheng Luo from Alibaba,
CET 12:00pm/PDT 3:00am/CDT 5:00am/Beijing 6:00pm - Poster Session (8 papers)
CET 1:00pm/PDT 4:00am/CDT 6:00am/Beijing 7:00pm - Two-hour Break
CET 3:00pm/PDT 6:00am/CDT 8:00am/Beijing 9:00pm - Contributed Talks Session 2
CET 4:15pm/PDT 7:15am/CDT 9:15am/Beijing 10:15pm - Invited talk 'Shopping Recommendations at Pinterest' by Jiufeng Zhou and Abhishek Tayal from Pinterest
CET 4:45pm/PDT 7:45am/CDT 9:45am/Beijing 10:45pm - 15 Minutes Break
CET 5:00pm/PDT 8:00am/CDT 10:00am/Beijing 11:00pm - Keynote speak by Prof Jiawei Han from UIUC, 45mins
CET 5:45pm/PDT 8:45am/CDT 10:45am/Beijing 11:45pm - Invited talk 'Domain specific knowledge graphs: challenges and opportunities' by Omar Alonso from Instacart
CET 6:15pm/PDT 9:15am/CDT 11:15am/Beijing 12:15am - Invited talk 'Integrating 3rd-Party Knowledge Graphs: Lessons from Adobe Stock' by Tracy King from Adobe
CET 6:45pm/PDT 9:45am/CDT 11:45am/Beijing 12:45am - Closing

Accepted Papers

Oral Presentations

• Robert Barton, Tal Neiman and Changhe Yuan: Graph Neural Networks for Inconsistent Cluster Detection in Incremental Entity Resolution
• Apoorva Balyan, Atul Singh, Praveen Suram, Deepak Arora and Varun Srivastava: USING PRODUCT META INFORMATION FOR BIAS REMOVAL IN E-COMMERCE GRID SEARCH
• Amine Dadoun, Raphaël Troncy, Michael Defoin Platel, Riccardo Petitti and Gerardo Ayala Solano: Optimizing Email Marketing Campaigns in the Airline Industry using Knowledge Graph Embeddings
• Yaxuan Wang, Hanqing Lu, Yunwen Xu, Rahul Goutam, Yiwei Song: QUEEN: Neural Query Rewriting in E-commerce
• Xu Liu, Congzhe Su, Amey Barapatre, Xiaoting Zhao, Diane Hu, Chu-Cheng Hsieh and Jingrui He: Interpretable Attribute-based Action-aware Bandits for Within-Session Personalization in E-commerce
• Alexander Brinkmann and Christian Bizer: Improving Hierarchical Product Classification using Domain-specific Language Modelling


• Paul Missault, Arnaud de Myttenaere, Andreas Radler and Pierre-Antoine Sondag: Addressing Cold Start With Dataset Transfer In E-Commerce Learning To Rank
• Shivam Sharma, Sachin Yadav, Praveen Suram, Deepak Arora and Varun Srivastava: A Novel Approach for Multi-Lingual Product Search in e-commerce System using Session Graph
• Hanqing Lu, Yunwen Xu, Qingyu Yin, Tianyu Cao, Boris Aleksandrovsky, Yiwei Song, Xianlong Fan and Bing Yin: Unsupervised Synonym Extraction for Document Enhancement in E-commerce Search
• Priya Gupta and Han Cuize: Aggregated Customer Engagement Model
• Ramasubramanian Balasubramanian, Venugopal Mani, Abhinav Mathur, Sushant Kumar and Kannan Achan: Variational Inference for Category Recommendation in E-Commerce platforms
• Lucia Yu, Ethan Benjamin, Congzhe Su, Yinlin Fu, Jon Eskreis-Winkler, Xiaoting Zhao and Diane Hu: Personalization in E-commerce Product Search by User-Centric Ranking
• Suhas Ranganath, Shibsankar Das, Sanjay Thillaivasan, Shipra Agarwal and Varun Srivastava: Grouping Search Results with Product Graphs in E-commerce Platforms
• Himanshu Jain, Mridul Katta, Kartik Kale, Suhas Ranganath and Varun Srivastava: LLETOR : Localised Learning to Rank

Keynote Speaker

Jiawei Han


Jiawei Han is the Abel Bliss Professor in the Department of Computer Science at the University of Illinois. He received his Ph.D. in Computer Sciences at University of Wisconsin in 1985. He worked as assistant professor in Northwestern University in 1986-1987 and as assistant, associate, full and university chair professor in Simon Fraser University in 1987-2001 before joining UIUC in 2001. He has been researching into data mining, information network analysis, and database systems, and their various applications, with over 600 publications. He has chaired or served on over 100 program committees of international conferences and workshops, including PC co-chair for KDD, SDM, and ICDM conferences, vice chair for ICDE and ICDM conferences, and Americas Coordinator for a VLDB conference. His book "Data Mining: Concepts and Techniques" (2nd ed., Morgan Kaufmann, 2006) has been popularly used as a textbook worldwide.

Invited Talks

1) Tracy Holloway King


Tracy Holloway King is currently a principal scientist in Adobe's Sensei and Search organization, focusing on natural language processing. She is trained in linguistics at Stanford and worked as a researcher and research manager in PARC's Natural Language Theory and Technology group. She then moved into applied research in Microsoft's Bing search team. She was a product manager and an engineering manager in the Search Science applied research team at eBay, focusing on eCommerce search across the eBay sites. At A9 and Amazon, she focused on natural language processing for product search query understanding and then on utilization for sponsored products. Currently she is a principal scientist at Adobe working on search and natural language processing.

Talk Title: Integrating 3rd-Party Knowledge Graphs: Lessons from Adobe Stock

Building knowledge graph (KG)-based features in an eCommerce site involves many steps from deciding whether to build the KG yourself or use an existing one to integrating the KG into the eCommerce architecture and running AB tests for KG-based features. In this talk, I present lessons learned from integrating a 3rd-party KG into Adobe Stock. Adobe Stock sells stock assets such as photos, illustrations, and videos. Customers can search for assets in over 20 languages. Adobe Stock wanted a KG with mappings to these languages to improve search relevance. Relevance issues included missing mappings from concepts to specific languages and cross-talk from text-based asset tags. For example, on the French site, the query "pain" (bread) returned assets showing people with injuries due to the English textual tag for pain. We also wanted the KG to help with recommendations and intelligent back-off for null results queries. Lesson learned: be clear on your top priority use case for the KG and evaluate against that. We did an initial evaluation to determine whether the 3rd-party KG would be appropriate and what it would entail to integrate it into our systems. Lesson learned: evaluation of KGs requires significant time around applied science and engineering. After licensing the KG, we built a platform to host it, an extracted sub-graph for use in Adobe Stock, tools for different views, and an integration pipeline for the search index and query. Lessons learned: build tooling first so that you can quickly debug issues; design for the long run, but prioritize ruthlessly for what to build first. After building the first small search indices, many of the issues we knew to expect from the evaluation were showing up: the named entity mappings were overapplying; there were missing and orphaned mappings with some languages missing mappings for a concept; the confidence scores on the mappings were crucial for removing edge cases. Lesson learned: Knowing that an issue will arise and planning in advance how to address it is important but there is still a lot of work to take those plans to production. After adjusting for the known issues, we then conducted an extensive search relevance evaluation to look for remaining issues and to determine whether we needed to retrain the ranker with the new KG. Lesson learned: repeated evaluation after changes is imperative to catch remaining issues. Overall, we believe we made the right decision to integrate a specialized 3rd-party KG into Adobe Stock: the KG integration into Adobe Stock will bring greatly improved search relevance to our customers, especially for languages other than English, French, and German, while providing a more modular architecture around concept mappings. But for future KG explorations we will pay careful attention to the lessons learned for a faster path to production.

2) Xusheng Luo


Xusheng Luo is currently a senior engineer in Alibaba Group's product search & recommendation team, focusing on enabling knowledge graph to improve e-commerce search & recommendation services. Before joining Alibaba Group, he earned master's degree of computer science at Shanghai Jiao Tong University. His research interests include knowledge engineering and natural language processing and published papers in conferences such as SIGMOD, EMNLP and AAAI. As a core contributor of AliCoCo, the large-scale e-commerce concept net assisting various bussiness in Alibaba, he will give a talk about the motivation of AliCoCo, its main technical challenges during construction and its applications in real industrial scenario.

3) Omar Alonso


Omar Alonso is a senior engineering manager at Instacart where he is currently working on knowledge graphs and large scale labeling. He holds a PhD from UC Davis.

Talk Title: Domain specific knowledge graphs: challenges and opportunities

There is a lot of interest in knowledge graphs as a rich structure that can be used in many applications like search, recommendation, and question-answering. While there are examples of knowledge graphs in industry and academia, there is little information on the practical aspects of construction and integration with other components. We describe some of the challenges and opportunities, and outline the current efforts at Instacart to build a knowledge graph about food and related topics to help users with their information needs.

4) Abhishek Tayal


Abhishek Tayal leads the Shopping Recommendations team at Pinterest - a multi disciplinary team of ML scientists and Engineers, helping Pinners create a life they love through an inspirational Shopping experience. The team leverages a vast array of ML technologies to deliver the best in class Recommendations experience. Before Pinterest, Abhishek was leading a team at Twitter - Cortex working on a variety of problems in the space of recommendation systems at Twitter like Embeddings algorithms/infra, advanced Candidate Generation algorithms etc. Previously, Abhishek worked with Tellapart, an ad tech startup (acquired by Twitter), and the Prime Video recommendations team at Amazon, where he led the development of the first-generation ML-based recommendation system for videos. He holds a master's degree from the University of Southern California in LA

Talk Title: Shopping Recommendations at Pinterest

More than 40% users come to Pinterest to find products. Pinterest Shopping is the key to help those users create a life they love. In this talk, we will first demonstrate how Pinterest inspires users about shopping ideas and helps them find the right products. Then, we will talk about a whole page optimization system we have built on a specific Pinterest shopping surface, product detail page. Finally, we will conclude the talk with the learnings from the whole page optimization.

Call For Papers

The Knowledge Management in E-Commerce Workshop welcomes submissions from both researchers and industry practitioners in knowledge discovery and applications for e-Commerce, including data cleaning and (unsupervised/weakly supervised) learning from noisy data, representation learning and embeddings, information extraction from text and graphs, user behavior modeling, and applications such as search, recommendation, advertising and QA. Full paper submissions (maximum 8 pages) are solicited in the form of research papers which propose new techniques and advances with industrial potential as well as papers from industry that describe practical applications and innovations in e-Commerce applications. Short papers (maximum 4 pages) describing case studies or work-in-progress are also solicited. Exceptionally well-argued position papers are also welcome.

In addition to short and long papers above, we welcome extended abstracts (1-2 pages) covering topical areas of research that are appropriate for the workshop. Authors of selected abstracts will be invited for oral presentations.

Top papers from the workshop will be selected for the June 2021 IEEE Data Engineering Bulletin Journal. See previous Journal since 1977 here:

Papers are solicited for the following set of non-exhaustive topics related to knowledge extraction, management and e-Commerce:

Theory, Algorithms and Methods:

• Extracting product knowledge and constructing Knowledge Graph from structured, semi-structured and unstructured data.
• Effective integration and learning of domain specifc human knowledge and human labels.
• Novel methodologies about evaluations and data curation/collection
• Unsupervised learning, weakly supervised learning from noisy data.
• Multilingual learning.
• Large-scale user behavioral graph mining.
• Data quality assessment for large scale product Knowledge Graphs.
• Infrastructure of Knowledge Graph-centric architectures.
• Novel definitions and theories regarding knowledge mining and representations.
• Deep Learning in knowledge extraction.
• Effective use of public Knowledge Graphs in e-Commerce.

Applications powered by Product Knowledge

• Product search, including query processing, mission understanding, products retrieval, ranking, and rendering.
• Product question answering
• Product recommendation
• Advertising
• User interfaces and visualization
• Experimental results using existing methods, including negative results of interest
• Systems issues in knowledge management in e-Commerce such as best practices, case studies, lessons learned, and feature descriptions

Vision, Opinion and Position Papers

We will also accept a small number of vision, opinion and position papers that provide discussions on challenges and roadmaps (for Knowledge Graph systems, applications and emerging models for e-commerce and product data).

Submission Directions

Submissions are limited to a total of eight (8) pages, including all content and references, and must be in PDF format and formatted according to the new Standard ACM Conference Proceedings Template. For LaTeX users: unzip, make, and use sample-sigconf.tex as a template.

Additional information about formatting and style files is available online at:

The accepted papers will be published online. Select papers may be invited to IEEE Data Engineering Bulletin Journal. Proceedings will be available for download after the conference.

All papers will be peer reviewed, single-blinded. Authors whose papers are accepted to the workshop will have the opportunity to participate in a poster session, and some set will be chosen for oral presentation.

We are using the EasyChair system for submissions. Please submit your paper using this link:

Please email any enquiries to