Diversified Multiple Instance Learning for Document-Level Multi-Aspect Sentiment Classification

Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing, pages 7012–7023, November 16–20, 2020. c⃝2020 Association for Computational Linguistics 7012 Diversified Multiple Instance Learning for Document-Level Multi-Aspect Sentiment Classification Yunjie Ji1,2, Hao Liu1, Bolei He1, Xinyan Xiao1∗, Hua Wu1, Yanhua Yu2 1Baidu Inc., Beijing, China 2Beijing University of Posts and Telecommunications {jiyunjie,liuhao24,hebolei,xiaoxinyan,wu hua}@baidu.com yuyanhua@bupt.edu.cn Abstract Neural Document-level Multi-aspect Senti- ment Classification (DMSC) usually requires a lot of manual aspect-level sentiment anno- tations, which is time-consuming and labo- rious. As document-level sentiment labeled data are widely available from online service, it is valuable to perform DMSC with such free document-level annotations. To this end, we propose a novel Diversified Multiple In- stance Learning Network (D-MILN), which is able to achieve aspect-level sentiment classifi- cation with only document-level weak supervi- sion. Specifically, we connect aspect-level and document-level sentiment by formulating this problem as multiple instance learning, provid- ing a way to learn aspect-level classifier from the back propagation of document-level super- vision. Two diversified regularizations are fur- ther introduced in order to avoid the overfit- ting on document-level signals during training. Diversified textual regularization encourages the classifier to select aspect-relevant snip- pets, and diversified sentimental regularization prevents the aspect-level sentiments from be- ing overly consistent with document-level sen- timent. Experimental results on TripAdvi- sor and BeerAdvocate datasets show that D- MILN remarkably outperforms recent weakly- supervised baselines, and is also comparable to the supervised method. 1 Introduction Document-level multi-aspect sentiment classifica- tion (DMSC) is a fine-grained sentiment analysis task, aiming to predict the sentiments of aspects in a document consisting of several sentences. In pre- vious studies, neural models have shown to be ef- fective for improving DMSC with the help of large amounts of aspect-level annotations (Chen et al., 2017; Xue and Li, 2018; Chen and Qian, 2019; ∗ corresponding author. + - A great location to stay at. The room was ordinary� the bathroom looked like an after thought, the shower was extremely small. Also the front desk clerks provided minimum service. Besides, the price is very expensive. Review overall: 2 Ratings room: 2 location: 4 service: 1 value: 1 A great location to stay at. The room was ordinary� the bathroom looked like an after thought, the shower was extremely small. Also the front desk clerks provided minimum service. Besides, the price is very expensive. Review overall: Sentiment room: location: service: value: A great location to stay at. The room was ordinary� the bathroom looked like an after thought, the shower was extremely small. Also the front desk clerks provided minimum service. Besides, the price is very expensive. Review overall: Ratings room: location: service: value: A great location to stay at. The room was ordinary� the bathroom looked like an after thought, the shower was extremely small. Also the front desk clerks provided minimum service. Besides, the price is very expensive. Review overall: Ratings room: location: service: value: 70% + - 30% 65 + - 70% 65 65 0.88 0.85 0.21 0.15 0.10 0.29 A great …

Content truncated to 500 words