- Title
- AOPSS: A Joint Learning Framework for Aspect-Opinion Pair Extraction as Semantic Segmentation
- Creator
- Wang, Chengwei; Peng, Tao; Zhang, Yue; Yue, Lin; Liu, Lu
- Relation
- 6th Asia-Pacific Web (APWeb) and Web-Age Information Management (WAIM) Joint Conference on Web and Big Data (APWeb-WAIM). Web and Big Data: 6th International Joint Conference, APWeb-WAIM 2022, Nanjing, China, November 25–27, 2022, Proceedings, Part II (Nanjing, China 25-27 November, 2022) p. 101-113
- Publisher Link
- http://dx.doi.org/10.1007/978-3-031-25198-6_8
- Publisher
- Springer
- Resource Type
- conference paper
- Date
- 2023
- Description
- Aspect-opinion pair extraction (AOPE) task, aiming at extracting aspect terms and their corresponding opinion terms in pairs, has caused widespread attention in recent years. Most studies focus on incorporating external knowledge, such as syntactic information. However, they are limited by the inadequate ability to capture long-distance information, and the utilization of external knowledge is more costly. In this paper, we propose AOPSS, a joint learning framework, to explore the AOPE task as semantic segmentation. As in most prior studies, we divide the AOPE task into two subtasks: entity recognition and relation detection. Specifically, AOPSS can synchronously capture task-invariant and task-specific features for the two subtasks without integrating any additional knowledge. Furthermore, we consider the interaction between entity and relation feature representations, which can improve the mutual heuristic effect for the two subtasks. Experimental results illustrate that our method achieves state-of-the-art performance on four public datasets, and we take further analysis to demonstrate the effectiveness of our approach.
- Subject
- sentiment analysis; relation extraction; entity recognition; semantic segmentation; joint learning
- Identifier
- http://hdl.handle.net/1959.13/1485611
- Identifier
- uon:51648
- Identifier
- ISBN:9783031251979
- Identifier
- ISSN:0302-9743
- Language
- eng
- Reviewed
- Hits: 1759
- Visitors: 1752
- Downloads: 0
| Thumbnail | File | Description | Size | Format |
|---|