{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/multi-pointer-co-attention-networks-for","title":"Multi-Pointer Co-Attention Networks for Recommendation","arxiv_id":"1801.09251","date":"2018-01-28","proceeding":null,"authors":["Yi Tay","Luu Anh Tuan","Siu Cheung Hui"],"abstract":"Many recent state-of-the-art recommender systems such as D-ATT, TransNet and\nDeepCoNN exploit reviews for representation learning. This paper proposes a new\nneural architecture for recommendation with reviews. Our model operates on a\nmulti-hierarchical paradigm and is based on the intuition that not all reviews\nare created equal, i.e., only a select few are important. The importance,\nhowever, should be dynamically inferred depending on the current target. To\nthis end, we propose a review-by-review pointer-based learning scheme that\nextracts important reviews, subsequently matching them in a word-by-word\nfashion. This enables not only the most informative reviews to be utilized for\nprediction but also a deeper word-level interaction. Our pointer-based method\noperates with a novel gumbel-softmax based pointer mechanism that enables the\nincorporation of discrete vectors within differentiable neural architectures.\nOur pointer mechanism is co-attentive in nature, learning pointers which are\nco-dependent on user-item relationships. Finally, we propose a multi-pointer\nlearning scheme that learns to combine multiple views of interactions between\nuser and item. Overall, we demonstrate the effectiveness of our proposed model\nvia extensive experiments on \\textbf{24} benchmark datasets from Amazon and\nYelp. Empirical results show that our approach significantly outperforms\nexisting state-of-the-art, with up to 19% and 71% relative improvement when\ncompared to TransNet and DeepCoNN respectively. We study the behavior of our\nmulti-pointer learning mechanism, shedding light on evidence aggregation\npatterns in review-based recommender systems.","url_abs":"http://arxiv.org/abs/1801.09251v2","url_pdf":"http://arxiv.org/pdf/1801.09251v2.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"multi-pointer-co-attention-networks-for","repo_url":"https://github.com/noveens/reviews4rec","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"multi-pointer-co-attention-networks-for","repo_url":"https://github.com/vanzytay/KDD2018_MPCN","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":null}],"tasks":[{"task_slug":"recommendation-systems","task_name":"Recommendation Systems"},{"task_slug":"representation-learning","task_name":"Representation Learning"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1801.09251","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}