Neural news recommendation with negative feedback

  • PDF / 1,416,507 Bytes
  • 11 Pages / 595.276 x 790.866 pts Page_size
  • 32 Downloads / 254 Views

DOWNLOAD

REPORT


REGULAR PAPER

Neural news recommendation with negative feedback Chuhan Wu1   · Fangzhao Wu2 · Yongfeng Huang1 · Xing Xie2 Received: 12 August 2020 / Accepted: 28 September 2020 / Published online: 7 October 2020 © China Computer Federation (CCF) 2020

Abstract News recommendation is important for online news services. Precise user interest modeling is critical for personalized news recommendation. Existing news recommendation methods usually rely on the implicit feedback of users like news clicks to model user interest. However, news click may not necessarily reflect user interests because users may click a news due to the attraction of its title but feel disappointed at its content. The dwell time of news reading is an important clue for user interest modeling, since short reading dwell time usually indicates low and even negative interest. Thus, incorporating the negative feedback inferred from the dwell time of news reading can improve the quality of user modeling. In this paper, we propose a neural news recommendation approach which can incorporate the implicit negative user feedback. We propose to distinguish positive and negative news clicks according to their reading dwell time, and respectively learn user representations from positive and negative news clicks via a combination of Transformer and additive attention network. In addition, we propose to compute a positive click score and a negative click score based on the relevance between candidate news representations and the user representations learned from the positive and negative news clicks. The final click score is a combination of positive and negative click scores. Besides, we propose an interactive news modeling method to consider the relatedness between title and body in news modeling. Extensive experiments on real-world dataset validate that our approach can achieve more accurate user interest modeling for news recommendation. Keywords  News recommendation · Dwell time · Negative feedback

1 Introduction Online news services such as Google News1 and Microsoft News2 can collect news from various sources and display them to users in a unified view (Das et al. 2007; Wu et al. 2020). However, a large number of news articles are generated every day and it is overwhelming for users to find their interested news (Okura et al. 2017). Thus, personalized news recommendation is critical for these news services to target * Chuhan Wu [email protected] Fangzhao Wu [email protected] Yongfeng Huang [email protected] Xing Xie [email protected] 1



Department of Electronic Engineering and BNRist, Tsinghua University, Beijing 100084, China



Microsoft Research Asia, Beijing 100080, China

2

13

Vol:.(1234567890)

user interests and alleviate information overload (Wu et al. 2019c). Accurate user interest modeling is a core problem in news recommendation. Existing news recommendation methods usually learn representations of users from their historical news click behaviors (Okura et al. 2017; Wang et al. 2018; Wu et al. 2019b, e). For example, Okura