Try This Instead: Personalized and Interpretable Substitute Recommendation

SIGIR '20: The 43rd International ACM SIGIR conference on research and development in Information Retrieval Virtual Event China July, 2020, pp. 891-900, 2020.

Cited by: 0|Bibtex|Views88|Links
EI
Keywords:
research questionsnormalized discounted cumulative gain at rank Kinterpretable substituteitem relationshiprecommender systemMore(12+)
Weibo:
We aim to answer the following research questions via experiments: RQ1: How effectively can A2CF perform personalized substitute recommendation compared with state-of-the-art baselines?

Abstract:

As a fundamental yet significant process in personalized recommendation, candidate generation and suggestion effectively help users spot the most suitable items for them. Consequently, identifying substitutable items that are interchangeable opens up new opportunities to refine the quality of generated candidates. When a user is browsing ...More

Code:

Data:

0
Introduction
  • On modern e-commerce platforms, it is a common practice to deploy recommender systems for retrieving items that match users’ personal interests [17].
  • In a variety of successful attempts, the identified complements can either stimulate further purchases after a user has bought a compatible item [14, 15], or attract potential customers via bundle advertisements beforehand [4, 54].
  • As a typical decision process in e-commerce [26], when a user is looking for a particular type of product to buy, she/he tends to first acquire a set of candidate items for comparison, and pick the most suitable one.
  • In a user’s decision-making process, recommending items that are substitutable and even superior to the one currently being browsed can expand the user’s view to make a better decision and eventually increase the chance of a successful purchase [50]
Highlights
  • On modern e-commerce platforms, it is a common practice to deploy recommender systems for retrieving items that match users’ personal interests [17]
  • We aim to answer the following research questions (RQs) via experiments: RQ1: How effectively can A2CF perform personalized substitute recommendation compared with state-of-the-art baselines?
  • RQ2: How is the quality of attribute-based interpretations generated by A2CF?
  • RQ4: How the hyperparameters affect the performance of A2CF in terms of recommendation effectiveness?
  • We study a new research problem, namely personalized and interpretable substitute recommendation, and propose a novel A2CF model as a solution
  • A2CF fully utilizes attribute information extracted from reviews, which effectively bridges user preferences and item properties for generating personalized substitute recommendations and endows the model with high explainability
Methods
  • The authors report the overall recommendation performance of A2CF with a unified set of parameters where d = 64, l = 1, γ = 0.7 and β =ε =d0.5 = 8.
  • The effect of different hyperparameter settings will be discussed in Section 6.4.
  • The authors use grid search to obtain their optimal hyperparameters.
Results
  • Following the settings in Section 5, the authors conduct experiments1 to evaluate the performance of A2CF regarding both recommendation effectiveness and interpretation quality.
  • The authors aim to answer the following research questions (RQs) via experiments: RQ1: How effectively can A2CF perform personalized substitute recommendation compared with state-of-the-art baselines?.
  • The first observation is that the proposed A2CF outperforms all baselines consistently and significantly (p-value < 0.01) on these three datasets.
  • By implementing substitution and personalization constraints via an attribute-aware scheme, A2CF represents the state-of-the-art effectiveness on this recommendation task
Conclusion
  • The authors study a new research problem, namely personalized and interpretable substitute recommendation, and propose a novel A2CF model as a solution.
  • A2CF fully utilizes attribute information extracted from reviews, which effectively bridges user preferences and item properties for generating personalized substitute recommendations and endows the model with high explainability.
  • The experimental results evidence that A2CF can yield superior performance on both recommendation and interpretation
Summary
  • Introduction:

    On modern e-commerce platforms, it is a common practice to deploy recommender systems for retrieving items that match users’ personal interests [17].
  • In a variety of successful attempts, the identified complements can either stimulate further purchases after a user has bought a compatible item [14, 15], or attract potential customers via bundle advertisements beforehand [4, 54].
  • As a typical decision process in e-commerce [26], when a user is looking for a particular type of product to buy, she/he tends to first acquire a set of candidate items for comparison, and pick the most suitable one.
  • In a user’s decision-making process, recommending items that are substitutable and even superior to the one currently being browsed can expand the user’s view to make a better decision and eventually increase the chance of a successful purchase [50]
  • Objectives:

    The authors aim to advance substitute recom- dation tasks [10, 12, 52].
  • In A2CF, with the user-attribute matrix X, the authors aim to learn the representations of each user ui and attribute an so that every scalar xin ∈ X can be inferred.
  • The authors aim to learn a triplet ranking function f (·, ·, ·) to generate scalar ranking scores such that f > f.
  • The authors aim to answer the following research questions (RQs) via experiments:
  • Methods:

    The authors report the overall recommendation performance of A2CF with a unified set of parameters where d = 64, l = 1, γ = 0.7 and β =ε =d0.5 = 8.
  • The effect of different hyperparameter settings will be discussed in Section 6.4.
  • The authors use grid search to obtain their optimal hyperparameters.
  • Results:

    Following the settings in Section 5, the authors conduct experiments1 to evaluate the performance of A2CF regarding both recommendation effectiveness and interpretation quality.
  • The authors aim to answer the following research questions (RQs) via experiments: RQ1: How effectively can A2CF perform personalized substitute recommendation compared with state-of-the-art baselines?.
  • The first observation is that the proposed A2CF outperforms all baselines consistently and significantly (p-value < 0.01) on these three datasets.
  • By implementing substitution and personalization constraints via an attribute-aware scheme, A2CF represents the state-of-the-art effectiveness on this recommendation task
  • Conclusion:

    The authors study a new research problem, namely personalized and interpretable substitute recommendation, and propose a novel A2CF model as a solution.
  • A2CF fully utilizes attribute information extracted from reviews, which effectively bridges user preferences and item properties for generating personalized substitute recommendations and endows the model with high explainability.
  • The experimental results evidence that A2CF can yield superior performance on both recommendation and interpretation
Tables
  • Table1: Statistics of datasets in use
  • Table2: Recommendation results. Numbers in bold face are the best results for corresponding metrics
  • Table3: Quantitative results on interpretation quality
  • Table4: Ablation test with different model architectures
Download tables as Excel
Related work
  • 7.1 Product Relationship Mining

    Understanding how products relate to each other is important to the fulfilment of customer satisfaction in different online shopping stages. In economics literature, two common product relationships are substitution and complement [23]. Recently, product relationship mining has become a prospective research direction to enhance existing recommender systems that are unable to differentiate the relationships among products. The problem of mining product relationships is first introduced to recommendation research in [24], where a topic model-based approach Sceptre is proposed. With latent Dirichlet allocation (LDA), Sceptre infers product relationships by comparing the topic distributions of two products. More recently, the trend of utilizing reviews has carried over to neural network approaches, such as RRN [49] that learns product embeddings from both the textual reviews and manually crafted features. In contrast to those hybrid models combining reviews with product graphs, pure graph-based methods like PMSC [43] and SPEM [50] are also introduced to fully utilize various properties in a product graph, e.g., path constraints and node proximities.
Funding
  • This work has been supported by Australian Research Council (Grant No DP190101985, DP170103954 and FT200100825) and National Natural Science Foundation of China (Grant No NSFC 61806035, U1936217, 61732008 and 61725203)
Reference
  • Jinze Bai, Chang Zhou, Junshuai Song, Xiaoru Qu, Weiting An, Zhao Li, and Jun Gao. 2019. Personalized Bundle List Recommendation. In WWW. 60–71.
    Google ScholarLocate open access versionFindings
  • Hongxu Chen, Hongzhi Yin, Tong Chen, Weiqing Wang, Xue Li, and Xia Hu. 2020. Social Boosted Recommendation with Folded Bipartite Network Embedding. TKDE (2020).
    Google ScholarLocate open access versionFindings
  • Hongxu Chen, Hongzhi Yin, Weiqing Wang, Hao Wang, Quoc Viet Hung Nguyen, and Xue Li. 2018. PME: projected metric embedding on heterogeneous networks for link prediction. In SIGKDD.
    Google ScholarFindings
  • Liang Chen, Yang Liu, Xiangnan He, Lianli Gao, and Zibin Zheng. 2019. Matching user with item set: collaborative bundle recommendation with deep attention network. In IJCAI. 2095–2101.
    Google ScholarFindings
  • Tong Chen, Hongzhi Yin, Hongxu Chen, Rui Yan, Quoc Viet Hung Nguyen, and Xue Li. 2019. AIR: Attentional intention-aware recommender systems. In ICDE. 304–315.
    Google ScholarLocate open access versionFindings
  • Tong Chen, Hongzhi Yin, Quoc Viet Hung Nguyen, Wen-Chih Peng, Xue Li, and Xiaofang Zhou. 2020. Sequence-Aware Factorization Machines for Temporal Predictive Analytics. ICDE (2020).
    Google ScholarLocate open access versionFindings
  • Zhiyong Cheng, Ying Ding, Xiangnan He, Lei Zhu, Xuemeng Song, and Mohan S Kankanhalli. 2018. A 3NCF: An Adaptive Aspect Attention Model for Rating Prediction.. In IJCAI. 3748–3754.
    Google ScholarFindings
  • Zhiyong Cheng, Ying Ding, Lei Zhu, and Mohan Kankanhalli. 201Aspectaware latent factor model: Rating prediction with ratings and reviews. In WWW. 639–648.
    Google ScholarLocate open access versionFindings
  • Gayatree Ganu, Yogesh Kakodkar, and AméLie Marian. 2013. Improving the quality of predictions using textual information in online user reviews. Information Systems 38, 1 (2013), 1–15.
    Google ScholarLocate open access versionFindings
  • Xinyu Guan, Zhiyong Cheng, Xiangnan He, Yongfeng Zhang, Zhibo Zhu, Qinke Peng, and Tat-Seng Chua. 2019. Attentive aspect modeling for review-aware recommendation. TOIS 37, 3 (2019), 28.
    Google ScholarLocate open access versionFindings
  • Lei Guo, Hongzhi Yin, Qinyong Wang, Tong Chen, Alexander Zhou, and Nguyen Quoc Viet Hung. 2019. Streaming session-based recommendation. In SIGKDD. 1569–1577.
    Google ScholarFindings
  • Xiangnan He, Tao Chen, Min-Yen Kan, and Xiao Chen. 2015. Trirank: Reviewaware explainable recommendation by modeling aspects. In CIKM. 1661–1670.
    Google ScholarFindings
  • Xiangnan He, Lizi Liao, Hanwang Zhang, Liqiang Nie, Xia Hu, and Tat-Seng Chua. 2017. Neural collaborative filtering. In WWW. 173–182.
    Google ScholarLocate open access versionFindings
  • Wang-Cheng Kang, Eric Kim, Jure Leskovec, Charles Rosenberg, and Julian McAuley. 2019. Complete the Look: Scene-based Complementary Product Recommendation. In CVPR. 10532–10541.
    Google ScholarFindings
  • Wang-Cheng Kang, Mengting Wan, and Julian McAuley. 2018. Recommendation through mixtures of heterogeneous item relationships. In CIKM. 1143–1152.
    Google ScholarFindings
  • Diederik P Kingma and Jimmy Ba. 2015. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980 (2015).
    Findings
  • Yehuda Koren, Robert Bell, and Chris Volinsky. 2009. Matrix factorization techniques for recommender systems. Computer (2009).
    Google ScholarLocate open access versionFindings
  • Pigi Kouki, Ilias Fountalis, Nikolaos Vasiloglou, Nian Yan, Unaiza Ahsan, Khalifeh Al Jadda, and Huiming Qu. 2019. Product collection recommendation in online retail. In RecSys. 486–490.
    Google ScholarLocate open access versionFindings
  • Yann LeCun, Yoshua Bengio, and Geoffrey Hinton. 2015. Deep learning. Nature 521, 7553 (2015), 436.
    Google ScholarLocate open access versionFindings
  • Binyang Li, Lanjun Zhou, Shi Feng, and Kam-Fai Wong. 2010. A unified graph model for sentence-based opinion retrieval. In ACL. 1367–1375.
    Google ScholarFindings
  • Huayu Li, Yong Ge, Richang Hong, and Hengshu Zhu. 2016. Point-of-interest recommendations: Learning potential check-ins from friends. In SIGKDD. 975– 984.
    Google ScholarLocate open access versionFindings
  • Qinghua Liu, Andrew Henry Reiner, Arnoldo Frigessi, and Ida Scheel. 2019. Diverse personalized recommendations with uncertainty from implicit preference data with the Bayesian Mallows model. Knowledge-Based Systems 186 (2019), 104960.
    Google ScholarLocate open access versionFindings
  • Andreu Mas-Colell, Michael Dennis Whinston, Jerry R Green, et al. 1995. Microeconomic theory. Vol. 1. Oxford university press New York.
    Google ScholarFindings
  • Julian McAuley, Rahul Pandey, and Jure Leskovec. 2015. Inferring networks of substitutable and complementary products. In SIGKDD. 785–794.
    Google ScholarLocate open access versionFindings
  • Tomas Mikolov, Kai Chen, Greg Corrado, and Jeffrey Dean. 2013. Efficient estimation of word representations in vector space. arXiv preprint arXiv:1301.3781 (2013).
    Findings
  • Wendy W Moe. 2006. An empirical two-stage choice model with varying decision rules applied to internet clickstream data. Journal of Marketing Research 43, 4 (2006), 680–692.
    Google ScholarLocate open access versionFindings
  • Apurva Pathak, Kshitiz Gupta, and Julian McAuley. 2017. Generating and personalizing bundle recommendations on steam. In SIGIR. 1073–1076.
    Google ScholarFindings
  • Vineeth Rakesh, Suhang Wang, Kai Shu, and Huan Liu. 2019. Linked variational autoencoders for inferring substitutable and supplementary items. In WSDM. 438–446.
    Google ScholarLocate open access versionFindings
  • Steffen Rendle, Christoph Freudenthaler, Zeno Gantner, and Lars Schmidt-Thieme. 2009. BPR: Bayesian personalized ranking from implicit feedback. In UAI. 452– 461.
    Google ScholarFindings
  • Xuemeng Song, Fuli Feng, Jinhuan Liu, Zekun Li, Liqiang Nie, and Jun Ma. 2017. Neurostylist: Neural compatibility modeling for clothing matching. In MM. 753– 761.
    Google ScholarLocate open access versionFindings
  • Nitish Srivastava, Geoffrey Hinton, Alex Krizhevsky, Ilya Sutskever, and Ruslan Salakhutdinov. 2014. Dropout: a simple way to prevent neural networks from overfitting. JMLR (2014), 1929–1958.
    Google ScholarLocate open access versionFindings
  • Ke Sun, Tong Chen, Tieyun Qian, Yiqi Chen, Hongzhi Yin, and Ling Chen. 2019. What can history tell us? Identifying relevant sessions for next-item recommendation. In CIKM. 1593–1602.
    Google ScholarFindings
  • Yu Sun, Nicholas Jing Yuan, Xing Xie, Kieran McDonald, and Rui Zhang. 2017. Collaborative intent prediction with real-time contextual data. TOIS (2017).
    Google ScholarLocate open access versionFindings
  • Jian Tang, Meng Qu, Mingzhe Wang, Ming Zhang, Jun Yan, and Qiaozhu Mei. 2015. Line: Large-scale information network embedding. In WWW. 1067–1077.
    Google ScholarFindings
  • Mengting Wan, Di Wang, Jie Liu, Paul Bennett, and Julian McAuley. 2018. Representing and Recommending Shopping Baskets with Complementarity, Compatibility and Loyalty. In CIKM. 1133–1142.
    Google ScholarFindings
  • Nan Wang, Hongning Wang, Yiling Jia, and Yue Yin. 2018. Explainable recommendation via multi-task learning in opinionated text data. In SIGIR. 165–174.
    Google ScholarFindings
  • Qinyong Wang, Hongzhi Yin, Tong Chen, Zi Huang, Hao Wang, Yanchang Zhao, and Nguyen Quoc Viet Hung. 2020. Next Point-of-Interest Recommendation on Resource-Constrained Mobile Devices. In The Web Conference. 906–916.
    Google ScholarLocate open access versionFindings
  • Weiqing Wang, Hongzhi Yin, Ling Chen, Yizhou Sun, Shazia Sadiq, and Xiaofang Zhou. 2015. Geo-SAGE: A geographical sparse additive generative model for spatial item recommendation. In SIGKDD. 1255–1264.
    Google ScholarFindings
  • Weiqing Wang, Hongzhi Yin, Zi Huang, Qinyong Wang, Xingzhong Du, and Quoc Viet Hung Nguyen. 2018. Streaming ranking based recommender systems. In SIGIR. 525–534.
    Google ScholarLocate open access versionFindings
  • Weiqing Wang, Hongzhi Yin, Shazia Sadiq, Ling Chen, Min Xie, and Xiaofang Zhou. 2016. Spore: A sequential personalized spatial item recommender system. In ICDE.
    Google ScholarFindings
  • Xiang Wang, Xiangnan He, Yixin Cao, Meng Liu, and Tat-Seng Chua. 2019. KGAT: Knowledge Graph Attention Network for Recommendation. SIGKDD (2019).
    Google ScholarLocate open access versionFindings
  • Xiang Wang, Xiangnan He, Meng Wang, Fuli Feng, and Tat-Seng Chua. 2019. Neural graph collaborative filtering. In SIGIR. 165–174.
    Google ScholarLocate open access versionFindings
  • Zihan Wang, Ziheng Jiang, Zhaochun Ren, Jiliang Tang, and Dawei Yin. 2018. A path-constrained framework for discriminating substitutable and complementary products in e-commerce. In WSDM. 619–627.
    Google ScholarLocate open access versionFindings
  • Xuchen Yao, Benjamin Van Durme, and Peter Clark. 2013. Automatic coupling of answer extraction and information retrieval. In ACL. 159–165.
    Google ScholarLocate open access versionFindings
  • Hongzhi Yin, Zhiting Hu, Xiaofang Zhou, Hao Wang, Kai Zheng, Quoc Viet Hung Nguyen, and Shazia Sadiq. 2016. Discovering interpretable geo-social communities for user behavior prediction. In ICDE. 942–953.
    Google ScholarLocate open access versionFindings
  • Hongzhi Yin, Xiaofang Zhou, Bin Cui, Hao Wang, Kai Zheng, and Quoc Viet Hung Nguyen. 2016. Adapting to user interest drift for poi recommendation. TKDE (2016), 2566–2581.
    Google ScholarLocate open access versionFindings
  • Hongzhi Yin, Xiaofang Zhou, Yingxia Shao, Hao Wang, and Shazia Sadiq. 2015. Joint modeling of user check-in behaviors for point-of-interest recommendation. In CIKM. 1631–1640.
    Google ScholarFindings
  • Lei Zhang and Bing Liu. 2014. Aspect and entity extraction for opinion mining. In Data mining and knowledge discovery for big data. 1–40.
    Google ScholarFindings
  • Mingyue Zhang, Xuan Wei, Xunhua Guo, Guoqing Chen, and Qiang Wei. 2019. Identifying Complements and Substitutes of Products: A Neural Network Framework Based on Product Embedding. TKDD 13, 3 (2019), 34.
    Google ScholarLocate open access versionFindings
  • Shijie Zhang, Hongzhi Yin, Qinyong Wang, Tong Chen, Hongxu Chen, and Quoc Viet Hung Nguyen. 2019. Inferring substitutable products with deep network embedding. IJCAI (2019), 4306–4312.
    Google ScholarLocate open access versionFindings
  • Yongfeng Zhang and Xu Chen. 2020. Explainable recommendation: A survey and new perspectives. Foundations and Trends in Information Retrieval (2020).
    Google ScholarLocate open access versionFindings
  • Yongfeng Zhang, Guokun Lai, Min Zhang, Yi Zhang, Yiqun Liu, and Shaoping Ma. 2014. Explicit factor models for explainable recommendation based on phrase-level sentiment analysis. In SIGIR. 83–92.
    Google ScholarLocate open access versionFindings
  • Yongfeng Zhang, Haochen Zhang, Min Zhang, Yiqun Liu, and Shaoping Ma. 2014. Do users rate or review?: Boost phrase-level sentiment labeling with review-level sentiment classification. In SIGIR. 1027–1030.
    Google ScholarFindings
  • Tao Zhu, Patrick Harrington, Junjun Li, and Lei Tang. 2014. Bundle recommendation in ecommerce. In SIGIR. 657–666.
    Google ScholarLocate open access versionFindings
Your rating :
0

 

Tags
Comments