This paper presents a novel way to learn Chinese polarity lexicons by using both external relations and internal formation of Chinese words, i.e. by integrating two kinds of different but complementary models: graph models and morphological feature-based models. The polarity detection is first treated as a semi-supervised learning in a graph, and then machine learning is used based on morphological features of Chinese words. The results show that the integration of morphological feature-based models and graph models significantly outperforms the baselines. Copyright © 2010 Springer-Verlag Berlin Heidelberg.
|Title of host publication||Information Retrieval Technology: 6th Asia Information Retrieval Societies Conference, AIRS 2010, Taipei, Taiwan, December 1-3, 2010. Proceedings|
|Editors||Pu-Jen CHENG , Min-Yen KAN , Wai LAM , Preslav NAKOV|
|Place of Publication||Berlin, Germany|
|ISBN (Print)||9783642171871, 3642171877, 9783642171864, 3642171869|
|Publication status||Published - 2010|
CitationLu, B., Song, Y., Zhang, X., & Tsou, B. K. (2010). Learning Chinese polarity lexicons by integration of graph models and morphological features. In P.-J. Cheng, M.-Y. Kan, W. Lam & P. Nakov (Eds.), Information Retrieval Technology: 6th Asia Information Retrieval Societies Conference, AIRS 2010, Taipei, Taiwan, December 1-3, 2010. Proceedings (pp. 466-477). Berlin, Germany: Springer-Verlag.
- Polarity lexicon induction
- Graph models
- Chinese morphology