AkihikoWatanabe / paper_notes

たまに追加される論文メモ
https://AkihikoWatanabe.github.io/paper_notes
15 stars 0 forks source link

Personalized Headline Generation with Enhanced User Interest Perception, Zhang+, ICANN'22 #928

Open AkihikoWatanabe opened 11 months ago

AkihikoWatanabe commented 11 months ago

https://link.springer.com/chapter/10.1007/978-3-031-15931-2_65

AkihikoWatanabe commented 11 months ago

Personalized news headline generation aims to model users interests through their news browsing history, and then generate news headlines based on their preferences. Using the user representation of existing news recommendation systems to personalize news headlines has been very successful, but there are still many problems to be solved. Firstly, news recommendation and headline generation tasks have different requirements for user interest modelling. Secondly, the user interest embedding approach does not consider the redundancy of user interests and the complementarity between user interests and candidate news texts. In this paper, we propose a new personalized headline generation framework to enhance the user interest perception. It can highlight user interests related to candidate texts for accurate embedding. We further offer to improve user interest representation using entity words in the news to meet the demand for news headline personalization. Extensive experiments show that our approach achieves better performance in headline generation tasks.

Translation (by gpt-3.5-turbo)