Beyond clicks: dwell time for personalization

By Xing Yi, Liangjie Hong, Erheng Zhong, Nathan Nan Liu, Suju Rajan. 2014

283 citations recorded by the source corpus.

Subfields: Information Systems, Management Science and Operations Research, Recommender Systems

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Information Systems

1. Xiangnan He
2. Tat-seng Chua
3. Xiang Wang
4. Fuli Feng
5. Julian McAuley
6. Liqiang Nie
7. Ed H. Chi
8. Yongfeng Zhang

In Management Science and Operations Research

1. Xiangnan He
2. Tat-seng Chua
3. Julian McAuley
4. Fuli Feng
5. Ed H. Chi
6. Zhe Zhao
7. Ruining He
8. Liqiang Nie

In Recommender Systems

1. Xiangnan He
2. Tat-seng Chua
3. Fuli Feng
4. Liqiang Nie
5. Julian McAuley
6. Ed H. Chi
7. Ruining He
8. Lichan Hong

Citations in the corpus

• Ups and Downs: Modeling the Visual Evolution of Fashion Trends with One-Class Collaborative Filtering (2016, methodology): “To cope with the non-linearity of fashion trends, we propose to automatically discover the important fashion ‘epochs’ each of which captures a separate set of prevailing visual decision factors at play.”
• VBPR: Visual Bayesian Personalized Ranking from Implicit Feedback (2015, background): “The ‘historical feedback’ used to train such systems may come in the form of explicit feedback such as star ratings, or implicit feedback such as purchase histories, bookmarks, browsing logs, search patterns, mouse activities etc. (Yi et al., 2014).”
• Deep Neural Networks for YouTube Recommendations (2016, background): “Ranking by click-through rate often promotes deceptive videos that the user does not complete (“clickbait”) whereas watch time better captures engagement [13, 25].”
• Improving Post-Click User Engagement on Native Ads via Survival Analysis (2016, methodology): “These measures have been used as proxies of post-click experience in online advertising and organic search, e.g., to improve the performance of ranking algorithms [18], as well as in recommender systems, e.g., to estimate the relevance of an item to a user [34].”
• Time-Sensitive Recommendation From Recurrent User Activities (2015, background): “) have been increasingly argued to be ineffective to represent real engagement of users due to the sparseness and nosiness of the data [26].”
• RecSys Challenge 2015: ensemble learning with categorical features (2015, methodology): “We have found it useful to calculate features based on what is called dwell time [6] in information retrieval.”
• Promoting Positive Post-Click Experience for In-Stream Yahoo Gemini Users (2015, background): “, to estimate the relevance of an item to a user [26].”
• Fairness in Recommendation Ranking through Pairwise Comparisons (2019, background): “inking for recommender systems. Part of the challenge in studying fairness in recommender systems is that they are complex. They often consist of multiple models [47, 24], must balance multiple goals [36, 50], and are di- cult to evaluate due to extreme and skewed sparsity [8] and numerous dynamics ”
• Returning is Believing: Optimizing Long-term User Engagement in Recommender Systems (2017, background): “For example, users may click on an item because of link bait, but may get disappointed and decide not to come back [28].”
• MeLU: Meta-Learned User Preference Estimator for Cold-Start Recommendation (2019, background): “The output layer, which is the subsequent layer of the decision-making layers, estimates user preferences that could be ratings, implicit feedback [7], or dwell times [29].”
• Towards a Fair Marketplace: Counterfactual Evaluation of the trade-off between Relevance, Fairness & Satisfaction in Recommendation Systems (2018, background): “It is well known that recommending relevant content has a positive impact on user satisfaction [28]; so we expect to achieve higher Industry and Case Study Paper CIKM’18, October 22-26, 2018, Torino, Italy”
• Denoising Implicit Feedback for Recommendation (2020, methodology): “and skip, and then incorporated a new rating function with weighted feedback into the recommendation. Besides, dwell time has been considered to be well correlated with the post-click user preference [18, 36]. Yi et al. [36] explored how to normalize dwell time, and then used them to train collabora”
• Learning to Rank (2005, unknown): “[51] – Clicks vs dwell time as an indicator of user satisfaction and engagement”