{"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/a-context-aware-user-item-representation","title":"A Context-Aware User-Item Representation Learning for Item Recommendation","arxiv_id":"1712.02342","date":"2017-12-29","proceeding":null,"authors":["Wu Libing","Quan Cong","Li Chenliang","Wang Qian","Zheng Bolong"],"abstract":"Both reviews and user-item interactions (i.e., rating scores) have been\nwidely adopted for user rating prediction. However, these existing techniques\nmainly extract the latent representations for users and items in an independent\nand static manner. That is, a single static feature vector is derived to encode\nher preference without considering the particular characteristics of each\ncandidate item. We argue that this static encoding scheme is difficult to fully\ncapture the users' preference. In this paper, we propose a novel context-aware\nuser-item representation learning model for rating prediction, named CARL.\nNamely, CARL derives a joint representation for a given user-item pair based on\ntheir individual latent features and latent feature interactions. Then, CARL\nadopts Factorization Machines to further model higher-order feature\ninteractions on the basis of the user-item pair for rating prediction.\nSpecifically, two separate learning components are devised in CARL to exploit\nreview data and interaction data respectively: review-based feature learning\nand interaction-based feature learning. In review-based learning component,\nwith convolution operations and attention mechanism, the relevant features for\na user-item pair are extracted by jointly considering their corresponding\nreviews. However, these features are only review-driven and may not be\ncomprehensive. Hence, interaction-based learning component further extracts\ncomplementary features from interaction data alone, also on the basis of\nuser-item pairs. The final rating score is then derived with a dynamic linear\nfusion mechanism. Experiments on five real-world datasets show that CARL\nachieves significantly better rating prediction accuracy than existing\nstate-of-the-art alternatives. Also, with attention mechanism, we show that the\nrelevant information in reviews can be highlighted to interpret the rating\nprediction.","url_abs":"http://arxiv.org/abs/1712.02342v5","url_pdf":"http://arxiv.org/pdf/1712.02342v5.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":"a-context-aware-user-item-representation","repo_url":"https://github.com/WHUIR/CARL","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"tf","reach":null}],"tasks":[{"task_slug":"prediction","task_name":"Prediction"},{"task_slug":"representation-learning","task_name":"Representation Learning"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}