Valence-arousal prediction of Chinese Words with multi-layer corpora
| Autor: | Hongjie Cen, Yuhong Xu, Jundong Wang, Peijie Huang, Xinrui Zhang, Jiecong Tang, Siyuan Chen, Qiangjia Huang, Piyuan Lin |
|---|---|
| Rok vydání: | 2016 |
| Předmět: |
Training set
business.industry Computer science Sentiment analysis 02 engineering and technology computer.software_genre Euclidean distance 030507 speech-language pathology & audiology 03 medical and health sciences Semantic similarity 020204 information systems Multiple time dimensions 0202 electrical engineering electronic engineering information engineering Word2vec Artificial intelligence Valence (psychology) 0305 other medical science business computer Multi layer Natural language processing |
| Zdroj: | IALP |
| Popis: | This paper presents our valence-arousal (VA) prediction method in the IALP 2016 Shared Task of Dimensional Sentiment Analysis for Chinese Words. Dimensional approach represents affective states as continuous numerical values in multiple dimensions, such as the VA space, thus allowing for more fine-grained sentiment analysis. For VA prediction, existing works usually selected similar seeds for an unseen word based on semantic similarity by using ontology or word2vec. However the semantic similarity is sometimes quite different from the similarity of valence/arousal. Therefore, this paper proposes a VA prediction method with multi-layer corpora to address such difference. In semantic layer, we get the top N most similar words by word2vec. Then, we screen the selected similar words based on the training data in VA layer. Finally, we use external corpora of affective polarity and intensity lexicons to make further filtering. Experimental results show that the proposed methods in this study to predict the value of VA yields good performance for Chinese words, especially in V dimension. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|