Towards Controllable and Personalized Review Generation

Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing, pages 3237–3245, Hong Kong, China, November 3–7, 2019. c⃝2019 Association for Computational Linguistics 3237 Towards Controllable and Personalized Review Generation Pan Li New York University Stern School of Business 44 West 4th Street, NY, NY, 10012 pli2@stern.nyu.edu Alexander Tuzhilin New York University Stern School of Business 44 West 4th Street, NY, NY, 10012 atuzhili@stern.nyu.edu Abstract In this paper, we propose a novel model RevGAN that automatically generates control- lable and personalized user reviews based on the arbitrarily given sentimental and stylistic information. RevGAN utilizes the combina- tion of three novel components, including self- attentive recursive autoencoders, conditional discriminators, and personalized decoders. We test its performance on the several real-world datasets, where our model significantly out- performs state-of-the-art generation models in terms of sentence quality, coherence, person- alization and human evaluations. We also em- pirically show that the generated reviews could not be easily distinguished from the organi- cally produced reviews and that they follow the same statistical linguistics laws. 1 Introduction With the ever increasing interests in user- generated reviews on online marketplace websites, such as Amazon, Yelp and TripAdvisor, it is nec- essary to provide a range of tools that would en- courage users to provide feedback in a more effi- cient and effective manner, as only a small frac- tion of users really take time to write their own reviews (Chen and Xie, 2008). Automatic review generation, for example, takes the product infor- mation and user behavior as input and generates user reviews following the arbitrarily given users’ sentiment designation and writing style personal- ized towards the specific product and user. Researchers have proposed various types of product review generation methods (Yao et al., 2017; Dong et al., 2017; Lipton et al., 2015; Rad- ford et al., 2017) and achieved great performance. However, they did not consider the inner hierar- chical word-sentence-paragraph structure within user reviews, thus making their generation results significantly limited in length and coherence. (Li et al., 2015; Zang and Wan, 2017) did include the hierarchical connection in their review generation model, but they did not address the problem of controllable and personalized review generation targeted at the specific product and user, which is essential for the usefulness of generated reviews. Most importantly, all the aforementioned gener- ative models did not include production descrip- tions in the generation process, thus their genera- tion results lack credibility and diversity. To address these problems, we propose a novel model RevGAN that automatically generates high- quality user reviews given the information of prod- uct descriptions, sentiment labels and users’ his- torical reviews. The proposed RevGAN model fol- lows a three-staged process: In Stage 1, we pro- pose to use Self-Attentive Recursive Autoencoder for mapping the discrete user reviews and prod- uct descriptions into continuous embeddings for the advantage of capturing the ‘’essence” of tex- tual information and the convenience for subse- quent optimization processes. In Stage …

Content truncated to 500 words