July 12, 2018, Ann Arbor, Michigan, USA

2018 Workshop on ExplainAble Recommendation and Search (EARS 2018)

Co-located with The 41st International ACM SIGIR Conference on Research and Development in Information Retrieval

About

The motivation of the workshop is to promote the research and application of Explainable Recommendation and Search, under the background of Explainable AI in a more general sense. Explainable recommendation and search attempt to develop models or methods that not only generate high-quality recommendation or search results, but also intuitive explanations of the results for users or system designers, which can help improve the system transparency, persuasiveness, trustworthiness, and effectiveness, etc.

In a broader sense, researchers in the whole artificial intelligence community have also realized the importance of Explainable AI, which aims to address a wide range of AI explainability problems in deep learning, computer vision, automatic driving systems, and natural language processing tasks. As an important branch of AI research, this highlights the importance our IR/RecSys community to address the explainability issues of various recommendation and search systems.

We welcome contributions of both long and short papers from a wide range of topics, including but not limited to explainable recommendation and search models, incorporating multi-modal information for explanation, evaluation of explainable recommendation and search, user study for explainable recommendation and search, etc. More topics are listed in the call for papers.

EARS'18 (co-located with SIGIR'18)
Michigan League
911 N. University
Ann Arbor, MI 48109-1265
Sponsers

Program

9:00 – 9:35 Research Keynote by Prof. Paul Resnick (University of Michigan)
Title: Survey Equivalence: An Information-theoretic Measure of Classifier Accuracy When the Ground Truth is Subjective
9:40 – 10:15 Industry Keynote by Dr. Qingsong Hua (Alibaba Inc.)
Title: Shakespeare of Alibaba: Practice of Intelligent Recommendation Reason Generation in Alibaba
10:30 – 11:00 Panel Discussions
Qingsong Hua (Alibaba Inc.), Matt Lease (University of Texas at Austin), Paul Resnick (University of Michigan), Mark Sanderson (RMIT University), Wlodek Zadrozny (UNC Charlotte)
11:00 – 12:00 Paper presentations
Session 1: Explainable Search
11:00 – 11:10 Explainable Information Retrieval using X-rays of Documents. Noriaki Kawamae (The University of Tokyo)
11:10 – 11:20 Explaining Credibility in News Articles using Cross-Referencing. Dimitrios Bountouridis (Delft University of Technology), Monica Marrero (Delft University of Technology), Nava Tintarev (Delft University of Technology), and Claudia Hauff (Delft University of Technology)
11:20 – 11:30 Posthoc Interpretability of Learning to Rank Models using Secondary Training Data. Jaspreet Singh (L3S Research Centre) and Avishek Anand (L3S Research Centre)
Session 2: Explainable Recommendation
11:30 – 11:40 Explainable Recommendation via Multi-Task Learning in Opinionated Text Data. Nan Wang (Shanghai Jiao Tong University), Hongning Wang (University of Virginia), Yiling Jia (University of Virginia) and Yue Yin (University of Virginia)
11:40 – 11:50 Correlated Topic Modelling via Householder Flow. Luyang Liu (Beijing Institute of technology), Heyan Huang (Beijing Institute of technology), and Yang Gao (Beijing Institute of technology)
11:50 – 12:00 Layer-wise Relevance Propagation for Explainable Recommendations. Homanga Bharadhwaj (Indian Institute of Technology Kanpur)

Research Keynote

Title: Survey Equivalence: An Information-theoretic Measure of Classifier Accuracy When the Ground Truth is Subjective

Abstract: Many classification tasks have no objective ground truth. Examples include: which content or explanation is "better" according to some community? is this comment toxic? what is the political leaning of this news article? The traditional modeling approach assumes each item has an objective true state that is perceived by humans with some random error. It fails to account for the fact that people have greater agreement on some items than others. I will describe an alternative model where the true state is a distribution over labels that raters from a specified population would assign to an item. This leads to information gain as a theoretically justified and computationally tractable measure of a classifier's quality, and an intuitive interpretation of information gain in terms of the sample size for a survey that would yield the same expected error rate.

Bio: Prof. Paul Resnick is the Michael D. Cohen Collegiate Professor of Information and Associate Dean for Research at the University of Michigan School of Information. He was a pioneer in the fields of recommender systems and reputation systems. He recently started the Center for Social Media Responsibility, which encourages and helps social media platforms to meet their public responsibilities.

Industry Keynote

Title: Shakespeare of Alibaba: Practice of Intelligent Recommendation Reason Generation in Alibaba

Abstract: Explainable recommendation and search is a very promising topic in both academia and industry in the recent years. There are many human generated recommendation reasons for products in Alibaba Taobao to improve user experience and to increase user stickiness. However, relying on human-generated content will result in low coverage, low quality stability, and high financial expenditure. With the rapid development of deep learning technology in NLP, especially in the nature language generation field, we tried natural language generation approach in recommendation reason generation and achieved good results. We created recommendation reasons for auction and auction list, which covered millions of product categories in Taobao e-commerce, and the generated explanations were used for large-scale real-world trasactions in "2017 Double 11 Shopping Festival" without any manual checking. Industry-level real system experiments show that it was very difficult to distinguish whether the explanations are machine-generated or manually-written, and the content generation can be controlled in multiple dimensions such as text style, text length, topics, etc. We will introduce our solution and technical details about generating free-text explanations in this keynote.

Bio: Dr. Qingsong Hua has been working in Alibaba search algorithm team since 2013 and has led a lot of projects about search relevance, quality, and conversion effectiveness. He is responsible for the intelligent recommendation reason generation project, which won the biggest Alibaba technical award in "2017 Double 11 Online Shopping Festival". He is now responsible for the overall international search algorithm team in Alibaba.

Accepted Papers

  1. Explaining Credibility in News Articles using Cross-Referencing.
    Dimitrios Bountouridis (Delft University of Technology), Monica Marrero (Delft University of Technology), Nava Tintarev (Delft University of Technology), and Claudia Hauff (Delft University of Technology)
  2. Explainable Information Retrieval using X-rays of Documents.
    Noriaki Kawamae (The University of Tokyo)
  3. Posthoc Interpretability of Learning to Rank Models using Secondary Training Data.
    Jaspreet Singh (L3S Research Centre) and Avishek Anand (L3S Research Centre)
  4. Layer-wise Relevance Propagation for Explainable Recommendations.
    Homanga Bharadhwaj (Indian Institute of Technology Kanpur)
  5. Correlated Topic Modelling via Householder Flow.
    Luyang Liu (Beijing Institute of technology), Heyan Huang (Beijing Institute of technology), and Yang Gao (Beijing Institute of technology)
  6. Explainable Recommendation via Multi-Task Learning in Opinionated Text Data (invited talk).
    Nan Wang (Shanghai Jiao Tong University), Hongning Wang (University of Virginia), Yiling Jia (University of Virginia) and Yue Yin (University of Virginia)
  7. User-Oriented Explanation for Intelligent Candidate Recommendation in Liepin.com.
    Roy Shan (Liepin.com), Xueyan Xu (Liepin.com) and Zhijie Li (Liepin.com)

Call for Papers

We welcome contributions of both long and short papers from a wide range of topics, including but not limited to the following topics of interest:

  1. New Models for Explainable Recommendation and Search
    • Explainable shallow models for recommendation and search
    • Explainable neural models for recommendation and search
    • Explainable sequential modeling
    • Explainable optimization algorithms and theories
    • Causal inference for explainable recommendation
  2. Using Different Information Sources for Explanation
    • Text-based modeling and explanation
    • Image-based modeling and explanation
    • Using knowledge-base for explanation
    • Audio/Video-based modeling and explanation
    • Integrating heterogenous information for explanation
  3. User Behavior Analysis and HCI for Explanation
    • Explanation and user satisfaction
    • Eye tracking and attention modeling
    • Mouse movement analysis
  4. New Types of Explanations for Search and Recommendation
    • Textual sentence explanations
    • Visual explanations
    • Statistic-based explanations
    • Aggregated explanations
    • Context-aware explanations
  5. Evaluation of Explainable Recommendation and Search
    • Offline evaluation measures and protocols
    • Online evaluation measures and protocols
    • User study for explanation evaluation
  6. Applications of Explainable Recommendation and Search
    • Explainable product search and recommendation
    • Explainable web search
    • Explainable social recommendation
    • Explainable news recommendation
    • Explainable point-of-interest recommendation
    • Explainable multi-media search and recommendation

PAPER SUBMISSION GUIDLINES

LONG PAPERS The maximum length is 9 pages (plus up to 1 page of references). Each accepted long paper will be presented in a plenary session. Each accepted long paper will also be allocated a presentation slot in a poster session to encourage discussion and follow up between authors and attendees.

SHORT PAPERS The maximum length is 4 pages (plus up to 1 page of references). Each accepted short paper will be presented in a spot-light session. Each accepted short paper will also be allocated a presentation slot in a poster session to encourage discussion and follow up between authors and attendees.

EARS 2018 submissions are double-blind. All submissions and reviews will be handled electronically. EARS 2018 submissions should be prepared according to the standard double-column ACM SIG proceedings format. Additional information about formatting and style files is available on the ACM website. Papers must be submitted to easychair at https://easychair.org/conferences/?conf=ears2018 by 23:59, AoE (Anywhere on Earth) on May 15th, 2018.

For inquires about the workshop and submissions, please email to ears2018@easychair.org

Important Days

All time are 23:59, AoE (Anywhere on Earth)
May 15, 2018: Submission due
May 31, 2018: Paper notification
June 20, 2018: Camera ready submission
July 12, 2018: Workshop day

Organization

Workshop Co-Chairs

Yongfeng Zhang (Rutgers University - New Brunswick, USA)
Yi Zhang (University of California Santa Cruz, USA)
Min Zhang (Tsinghua University, Beijing, China)

Program Committee Members

Behnoush Abdollahi (University of Louisville)
Qingyao Ai (University of Massachusetts Amherst)
Rose Catherine (Carnegie Mellon University)
Xu Chen (Tsinghua University)
Michael Ekstrand (Boise State University)
Ruining He (Pinterest)
Xiangnan He (National University of Singapore)
Bart Knijnenburg (Clemson University)
Aonghus Lawlor (University College Dublin)
Piji Li (The Chinese University of Hong Kong)
Julian Mcauley (University of California San Diego)
Sole Pera (Boise State University)
Zhaochun Ren (JD Data Science Lab)
Sungyong Seo (University of Southern California)
Xiang Wang (National University of Singapore)
Yao Wu (Twitter Inc.)
Hamed Zamani (University of Massachusetts Amherst)

THE VENUE

EARS'18 will be co-located with The 41st International ACM SIGIR Conference on Research and Development in Information Retrieval, to be held at Ann Arbor, Michigan. The conference venue is Michigan League, located at 911 N. University Ave, Ann Arbor, MI 48109-1265.