Personalized Transformer for Explainable Recommendation

Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing, pages 4947–4957 August 1–6, 2021. ©2021 Association for Computational Linguistics 4947 Personalized Transformer for Explainable Recommendation Lei Li1 Yongfeng Zhang2 Li Chen1 1Hong Kong Baptist University, Hong Kong, China 2Rutgers University, New Brunswick, USA 1{csleili,lichen}@comp.hkbu.edu.hk 2yongfeng.zhang@rutgers.edu Abstract Personalization of natural language generation plays a vital role in a large spectrum of tasks, such as explainable recommendation, review summarization and dialog systems. In these tasks, user and item IDs are important identi- fiers for personalization. Transformer, which is demonstrated with strong language model- ing capability, however, is not personalized and fails to make use of the user and item IDs since the ID tokens are not even in the same semantic space as the words. To address this problem, we present a PErsonalized Trans- former for Explainable Recommendation (PE- TER1), on which we design a simple and ef- fective learning objective that utilizes the IDs to predict the words in the target explanation, so as to endow the IDs with linguistic mean- ings and to achieve personalized Transformer. Besides generating explanations, PETER can also make recommendations, which makes it a unified model for the whole recommendation- explanation pipeline. Extensive experiments show that our small unpretrained model outper- forms fine-tuned BERT on the generation task, in terms of both effectiveness and efficiency, which highlights the importance and the nice utility of our design. 1 Introduction Recent years have witnessed the successful appli- cation of natural language generation. Many of the applications in fact require certain degree of per- sonalization, such as explainable recommendation (Zhang et al., 2014; Li et al., 2020c; Zhang and Chen, 2020), review generation (Dong et al., 2017), review summarization (Li et al., 2019), and conver- sational systems (Zhang et al., 2018; Chen et al., 2020). In these tasks, user and item IDs that distin- guish one user/item from the others are crucial to 1https://github.com/lileipisces/PETER personalization. For example, in recommender sys- tems, different users may care about different item features (e.g., style vs. quality), and different items may have different characteristics (e.g., fashionable vs. comfortable). The goal of explainable recom- mendation (Zhang and Chen, 2020) is to provide an explanation to a user for a recommended item, so as to justify how the recommendation might match his/her interests. That is, given a pair of user ID and item ID, the system needs to generate an explana- tion, such as “the style of the jacket is fashionable” (see the last column of Table 4 for more examples). Transformer (Vaswani et al., 2017), whose strong language modeling ability has been demon- strated on a variety of tasks (Radford et al., 2018; Devlin et al., 2019; Brown et al., 2020), however, is relatively under-explored for personalized natural language generation. Since IDs and words are in very different semantic spaces, it would be prob- lematic to directly put them together for attention learning, because by doing so, the IDs are treated …

Content truncated to 500 words