抄録
Sentiment (i.e., bad or good opinion) described in texts has been studied widely, and at three different levels: word, sentence, and document level. This paper describes a well-founded approach for the task of sentence level sentiment analysis by studying the relationship between sentiments conveyed through texts and structure of natural language by a method of numerical analysis. Different approaches have been employed to "sense" sentiment, especially from the texts, but none of those ever considered the valence based appraisal structure of sentiments which we have employed. Therefore the paper describes an approach to sense sentiments contained in a sentence by applying a numerical-valence based analysis. To meet this objective a linguistic tool, SenseNet, has been developed that provides lexical-units on the basis of each semantic verb frame obtained from the input sentence; assigns a numerical value to those based on their sense affinity; assesses the values using rules; and finally outputs sense-valence for each input sentence. Several experiments with a variety of datasets containing data from different domains have been conducted. The obtained results indicate significant performance gains over existing state-of-the-art approaches.
本文言語 | English |
---|---|
ホスト出版物のタイトル | 2007 10th International Conference on Computer and Information Technology, ICCIT |
DOI | |
出版ステータス | Published - 2007 |
外部発表 | はい |
イベント | 2007 10th International Conference on Computer and Information Technology, ICCIT - Dhaka 継続期間: 2007 12月 27 → 2007 12月 29 |
Other
Other | 2007 10th International Conference on Computer and Information Technology, ICCIT |
---|---|
City | Dhaka |
Period | 07/12/27 → 07/12/29 |
ASJC Scopus subject areas
- コンピュータ サイエンスの応用
- ソフトウェア