Fine-grained opinion mining (also called aspect-based sentiment analysis) aims at extracting knowledge about opinion targets (aspects), opinion holders and the opinions/sentiments expressed towards them, leading to structured opinion summaries. This task proves to be more crucial and challenging by providing a thorough analysis of an opinionated text, yet is under-discussed in the community compared to overall sentiment score classification. This tutorial aims at reviewing existing works in this field that covers 3 main subtasks, namely aspect-based sentiment classification, aspect-related extraction and summarization. We introduce various model constructions, including feature-based, rule-based and deep-learning-based models, that focus on exploiting complex word-level interactions among an input text and promote the generality of these methods to be adopted for efficient knowledge extraction. Besides single-domain studies, a further step is to explore cross-domain, cross-lingual and multi-modal strategies. Although much more challenging, these alternatives promote the development of fine-grained opinion mining, because there are only limited resources available with fine-grained annotations in real industries. We introduce a few existing studies and aim to provide more insights towards these cutting-edge research directions.
Wenya Wang: A research fellow in School of Computer Science and Engineering, Nanyang Technological University, Singapore. Her research interests include deep learning in Natural Language Processing, sentiment analysis and information extraction.
Jianfei Yu: A research scientist in School of Information Systems, Singapore Management University, Singapore. His research centers around deep learning and transfer learning in many Natural Language Processing tasks including sentiment analysis, information extraction, and question answering.
Sinno Jialin Pan: An Associate Professor at the School of Computer Science and Engineering at Nanyang Technological University (NTU), Singapore. His research interests include transfer learning, and its applications to sentiment analysis, wireless sensor networks, software engineering, etc.
Jing Jiang: An Associate Professor of Information Systems at Singapore Management University (SMU), Singapore. Her research interests include information extraction, sentiment analysis, question answering, text mining and social media analysis.
- Introduction to fine-grained opinion analysis
- Aspect-level sentiment classification
- Aspect extraction
- Opinion summarization
- Going further