Conference Proceeding Article
Opinion target extractionis a fundamental task in opinion mining. In recent years,neural network based supervised learning methods haveachieved competitive performance on this task. However, aswith any supervised learning method, neural network basedmethods for this task cannot work well when the training datacomes from a different domain than the test data. On the otherhand, some rule-based unsupervisedmethods have shown to berobust when applied to different domains. In this work, weuse rule-based unsupervised methods to create auxiliary labelsand use neural network models to learn a hiddenrepresentation that works well for different domains. When this hiddenrepresentation is used for opinion target extraction, we findthat it can outperform a number of strong baselines with alarge margin.
OS and Networks
Information Systems and Management
Proceedings of the 31st AAAI Conference on Artificial Intelligence: San Francisco, USA, 2017 February 4
Association for the Advancement of Artificial Intelligence
City or Country
San Francisco, USA
DING YING; YU, JIANFEI; and Jing JIANG.
Recurrent neural networks with auxiliary labels for cross-domain opinion target extraction. (2017). Proceedings of the 31st AAAI Conference on Artificial Intelligence: San Francisco, USA, 2017 February 4. Research Collection School Of Information Systems.
Available at: http://ink.library.smu.edu.sg/sis_research/3530
Creative Commons License
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License.