Oct 21, 2022

Deep Learning for Search Recommendation (DL4SR 2022)

Workshop held in conjunction with CIKM 2022

In the current digital world, web search engines and recommendation systems are continuously evolving, opening up new potential challenges every day which require more sophisticated and efficient data mining and machine learning solutions to satisfy the needs of sellers and consumers as well as marketers. The quality of search and recommendation systems impacts customer retention, time on site, and sales volume. For instance, with often sparse conversion rates, highly personalized contents, heterogeneous digital sources, more rigorous and effective models are required to be developed by research engineers and data scientists. At the same time, deep learning has started to show great impact in many industrial applications which are capable of processing complicated, large-scale and real-time data. Deep learning not only provides more opportunities to increase conversion rates and improve revenue through a positive customer experience, but also provides customers with personalized contents along with their personal shopping journey. Due to this rapid growth of the digital world, there is a need to bring professionals together from both academic research and the industry to solve real-world problems. This is exactly what this workshop aims to achieve. Topics of this workshop include deep learning based query understanding, personalization, representation learning, product retrieval, recommendation algorithm, ranking algorithms, etc.

We invite quality research contributions, industrial achievement addressing relevant deep learning challenges in the domain of search, recommendation and personalization. We invite submission of long and short papers of two to eight pages (including references), representing original research, preliminary research results, proposals for new work in academic or industry. All submissions will be single-blind and will be peer reviewed by an international program committee of researchers/industrial professionals with high reputation. Accepted submissions will be required to be presented at the workshop.

Topics of interest on Deep Learning approaches for eCommerce and marketing include, but are not limited to:

- Deep learning based search models

- Deep learning based recommendation models

- Deep learning based recommendation optimization models (e.g. deep reinforcement learning etc.)

- Recommendation and Personalization: user historical behavior-based, content-based etc.

- Representations Learning: various deep representations of products, queries, and customers including knowledge graph, embedding etc.

- Retrieval models and ranking (e.g., ranking algorithms, learning to rank, NLP models, retrieval models, etc).

- Query Understanding: Query intent, Query correction, Query suggestion, Query expansion, multi-modalities queries as well as Query embedding, classification etc

- Privacy issues in search and/or recommendation models

- Multimodal search and/or recommendation models

- Heterogeneous data analysis in search and/or recommendation models

- Industrial domain-specific applications of search, recommendation models

- Improved model for customer engagement in marketing etc.

Call for Papers

Submission link: https://easychair.org/conferences/?conf=dl4sr22

Paper Submission Deadline: August 15, 2022, 11:59 PM AoE.

Paper Notification: September 15, 2022, 11:59 PM AoE.

Camera Ready Version: September 30, 2022, 11:59 PM AoE.

Full-Day Workshop: October 21, 2022

This workshop follows the submission requirements of the CIKM conference.


- Long papers are up to 8 pages and short papers are up to 4 pages. These page limits include the bibliography and any possible appendices.

- Single-blind peer review

- All papers must be formatted according to the ACM sigconf template manuscript style, following the submission guidelines available at: https://www.acm.org/publications/proceedings-template.

- Papers should be submitted in PDF format, electronically, using the EasyChair submission system.

Publication: Accepted papers in this workshop may be included in the DL4SR 2022 Workshop Proceedings published online by CEUR.

Keynote Speakers


Beyond Being Accurate: Neural Modeling and Reinforcement Learning for Large-Scale Real-World Recommendation Problems

Ed H. Chi

Distinguished Scientist at Google

Fundamental improvements in recommendation and ranking have been much harder to come by, when compared with recent progress on other long-standing AI problems such as visual/audio machine perception and machine translation. Some reasons include: (1) large amounts of data making training difficult, yet having (2) noisy and sparse labels; (3) changing dynamics of context such as user preferences and items; and (4) low-latency requirement for a system response. Moreover, one new challenge is devising approaches to (5) learning more inclusive and robust models.

In this talk, Ed will touch upon many recent advances in neural modeling techniques for recommendations and their impact in Google products covering ~600 improvements over the last 5 years across these problems, including:
  • (a). Multi-task models with gated mixture of experts and utilizing TPUs for large sparse models
  • (b). Large output item spaces with Neural Deep Retrieval
  • (c). Policy gradient RL techniques with off-policy correction in recommendation models
  • (d). Diversification and slate optimization
  • (e). Adversarial approaches for inclusiveness and robustness for Classifiers and Recommenders

Speaker Bio:
Ed H. Chi is a Distinguished Scientist at Google, leading several machine learning research teams focusing on neural modeling, reinforcement learning, dialog chatbot models called LaMDA, reliable/robust machine learning, and recommendation systems in Google Brain team. His team has delivered significant improvements for YouTube, News, Ads, Google Play Store at Google with ~600 product improvements since 2013. With 39 patents and >150 research articles, he is also known for research on user behavior in web and social media. Prior to Google, he was the Area Manager and a Principal Scientist at Palo Alto Research Center's Augmented Social Cognition Group, where he led the team in understanding how social systems help groups of people to remember, think and reason. Ed completed his three degrees (B.S., M.S., and Ph.D.) in 6.5 years from University of Minnesota. Recognized as an ACM Distinguished Scientist and elected into the CHI Academy, he recently received a 20-year Test of Time award for research in information visualization. He has been featured and quoted in the press, including the Economist, Time Magazine, LA Times, and the Associated Press. An avid swimmer, photographer and snowboarder in his spare time, he also has a blackbelt in Taekwondo.


A Path Towards More Universal Question Answering Systems

Caiming Xiong

VP of AI Research at Salesforce

With the growing capabilities in how we represent and process structured and unstructured text, there are many opportunities to innovate newer and deeper search paradigms. This would allow us to go beyond building one-off models, each focusing on a specific question-answering (QA) task targeting a specific domain, reasoning skill, or answer type. In this talk, Caiming will start with a series of in-depth presentations on a few of the recent QA research conducted by the Salesforce AI Research team addressing several key questions in the QA domain.

Speaker Bio:
Caiming Xiong is Vice President of AI Research at Salesforce. He received his Ph.D. in the department of Computer Science and Engineering, University of Buffalo, SUNY and worked as a Postdoctoral Researcher Scholar at the University of California, Los Angeles. His research interests fall in the disciplines of ML/DL, natural language processing, multimedia, and AI for Good. He has served on the organizing committee of multiple workshops and has been conference area chair in top-tier conferences such as NeurIPS, ICLR, EMNLP, AC, AAAI etc.


Responsible Recommendation of Information Items: Veracity, Fairness, and Trust

Xiuzhen (Jenny) Zhang

Professor of RMIT University, Australia

Recommender systems have grown beyond E-commerce applications to novel domains. In particular, information items such as news articles, social media posts and reviews are complex items associated with properties such as author’s viewpoints, stance, reputation, credibility, and bias. We posit that the objective of recommendation of such complex information items shall go beyond high accuracy of satisfying user personal interests to high social responsibility. Socially responsible recommender systems call for shift to a new paradigm of data preparation, recommendation model and evaluation. In this talk, I will discuss our recent work on recommendation of social media items for the mitigation of misinformation and fairness of information sources, as well as user trust of information items.

Speaker Bio:
Xiuzhen (Jenny) Zhang is Professor of Data Science at RMIT University, Australia. Her research interests are data mining and machine learning, with a focus on textual data and social media data. She is especially interested in data science for social good, in areas such as misinformation detection and mitigation, law enforcement, and digital health. She has published over 100 papers in these areas. She is an associate editor of the journal Information Processing and Management, and has served on the organising committee of international and Australian conferences such as KDD, PAKDD, EMNLP, IEEE DSAA, ALTW, ADMA.


Scaling, Strengthening and Serving: Innovating through Large-Scale Deep Representations in E-commerce Product Search and Recommendation Applications

Belinda Zeng

GM, Head of Applied Science and Engineering - Amazon Search Science and AI

Recent work in e-commerce product search and recommendations have shown result improvement by using large scale pre-trained model based Deep Learning approach. However, this has not been an easy task in the real-world development and deployment due to several constraints: 1) scaling the size of model and training data with computational efficiency is challenging; 2) boosting the incumbent ML applications performance using large scale pre-trained model in the real business setting requires careful algorithm design and data strategy; and 3) serving millions of requests per second at high throughput and low latency is a daunting task. In this talk, I will share our recent work that addresses these issues and a vision for future innovations.

Speaker Bio:
Belinda Zeng is currently Head of Applied Science and Engineering in Amazon Search Science and AI. Over her career, she has been taking various leadership roles at Amazon (Alexa AI and Amazon Consumer Payments), Nielsen, Discover financial and Deloitte Consulting. She is currently leading Amazon's large scale AI program where she works with a group of top-notch scientists and engineers building universal semantic representations of e-commerce specific entities. This effort aims to bring Amazon services beyond the current state-of-the-art and unlock many new downstream applications that bring delightful experiences to Amazon customers. She obtained her PhD in Economics and Master in Mathematics in Indiana University - Bloomington.


A Deep Learning Virtuous Cycle for User Understanding

Jonathan Purnell

VP of Data Science at Spectrum Labs

Platforms can provide users with a wide variety of options which may be both exciting and overwhelming. While a recommendation system provides users with relevant recommendations balanced with exploring new categories, they also need to be wary not only of irrelevant recommendations but also potentially inappropriate suggestions. Particularly where third party vendors or user-generated content is involved, it’s paramount that recommendation systems extract sufficient insights to provide a tailored positive user experience. In this talk, we’ll explore the challenges of defining this problem (eg. appropriate vs inappropriate), building robust models, as well as an iterative workflow to adapt to evolving language. The overall theory will be supplemented with real-world use cases where being proactive about these challenges not improved user experiences but led to new marketing opportunities.

Speaker Bio:
Jonathan Purnell is the VP of Data Science at Spectrum Labs, where NLP is applied to help platforms recognize and respond to toxicity in user-generated content. He and his team have developed solutions to deliver actionable insights into behaviors such as hate speech and bullying over various forms of conversational text and across several languages. Formerly, as a Data Scientist for Krux and Salesforce DMP, Jon delivered Internet-scale distributed products using innovative distributed ML techniques. Before that, he was an Applied Scientist at Bing Ads and a collaborative researcher with BBN Technologies (a division of Raytheon). He holds a Ph.D. in computer science focused on machine learning.

Accepted Papers

- Yingyi Zhang, Xianneng Li, Yahe Yu, Jian Tang, Huanfang Deng, Junya Lu, Yeyin Zhang, Qiancheng Jiang, Yunsen Xian and Liqian Yu. Entire Cost Enhanced Multi-Task Model for Online-to-Offline Conversion Rate Prediction

- Gautam Kumar and Chikara Hashimoto. A Two-Phased Approach to Training Data Generation for Shopping Query Intent Prediction

- Vito Walter Anelli, Yashar Deldjoo, Tommaso Di Noia, Eugenio Di Sciascio, Antonio Ferrara, Daniele Malitesta and Claudio Pomo. Reshaping Graph Recommendation with Edge Graph Collaborative Filtering and Customer Reviews

- Debanjan Mahata, Navneet Agarwal, Dibya Gautam, Amardeep Kumar, Swapnil Parekh, Yaman Kumar Singla, Anish Acharya and Rajiv Ratn Shah. LDKP - A Dataset for Identifying Keyphrases from Long Scientific Documents

- Osayande P. Omondiagbe, Sherlock A Licorish and Stephen G. MacDonell. Preventing Negative Transfer on Sentiment Analysis in Deep Transfer Learning

- Tesi Xiao, Xia Xiao, Ming Chen and Youlong Chen. Field-wise Embedding Size Search via Structural Hard Auxiliary Mask Pruning for Click-Through Rate Prediction

- Yanchao Tan, Carl Yang, Xiangyu Wei, Ziyue Wu, Weiming Liu and Xiaolin Zheng. Partial Relaxed Optimal Transport for Denoised Recommendation

- Shaochuan Lin, Yicong Yu, Xiyu Ji, Taotao Zhou, Hengxu He, Zisen Sang, Jia Jia, Guodong Cao and Ning Hu. Spatiotemporal-Enhanced Network for Click-Through Rate Prediction in Location-based Services

- Tajuddeen Gwadabe, Mohammed Ali Al-Hababi and Ying Liu. Session-based Recommendation with Dual Graph Networks

- Qian Yu, Xiangdong Wu, Chen Yang, Zihao Zhao, Haoxin Liu, Chaosheng Fan, Changping Peng, Zhangang Lin, Jinghe Hu and Jingping Shao. Exploiting Global Behavior Contextual Correlation in Sequential Recommendation Augmentation

- Xue Li, Wei Shen and Denis Charles. TEDL: A Two-stage Evidential Deep Learning Method for Classification Uncertainty Quantification

- Xianjing Liu, Behzad Golshan, Kenny Leung, Aman Saini, Vivek Kulkarni and Ali Mollahosseini. TWICE - Twitter Content Embeddings



Wei Liu

University of Technology Sydney

Kexin Xie


Linsey Pang


James Bailey

The University of Melbourne

Longbing Cao

University of Technology Sydney

Yuxi Zhang


Contact: dlsr.workshop@gmail.com

Program Committee

Philippe Laban, Salesforce

Zhensong Qian, Amazon

Donglin Hu , Salesforce

Xinghao Yang, University of Technology Sydney

Ali Braytee, University of Technology, Sydney

Aneesh Chivukula, BITS Pilani

Mingze Ni, University of Technology, Sydney

Vito Ostuni, Netflix

Xianjing Liu, Twitter Inc.

Xue Li, Microsoft

Raghavendra Chalapathy, Walmart Labs