Beijing Key Lab of Intelligent Telecommunication Software and Multimedia, School of Computer Science, Beijing University of Posts and Telecommunications, China
Auto-generated profile, based on publicly available information
Yixiang Huang
Beijing Key Lab of Intelligent Telecommunication Software and Multimedia, School of Computer Science, Beijing University of Posts and Telecommunications, China
Sentiment analysis of user reviews has attracted increasing research attention due to its great underlying value for business. With the scarcity of public datasets, the majority of research is done on datasets with respect to social media and restaurants. To address this problem and advance relevant research in more fields, we present a large-scale Chinese tobacco review dataset collected from a leading tobacco forum in China. Besides an overall rating, we annotate each review from 6 different pre-defined aspects and we hope that the release of the dataset could shed some light on the research in the tobacco and sentiment analysis field. Aspect category sentiment analysis (ACSA) is an essential task to detect fine-grained sentiment polarities within a text. In order to further facilitate the ACSA research, we propose the intuitive yet effective TVBART model which combines the strengths of recent NLP advances and the voting method in ensemble learning. Experiments on our new dataset and the public ASAP dataset reveal that the proposed model outperforms state-of-the-art baselines on the ACSA task.