DocumentCode
2052450
Title
Contextual Analysis Processing Methods Able to Interpret Sentiments Evaluation Representations
Author
Itoh, Miho
Author_Institution
Keio Univ., Yokohama, Japan
fYear
2009
fDate
14-16 Sept. 2009
Firstpage
71
Lastpage
76
Abstract
A contextual analysis processing technique, consisting in determining the context understanding together with coherences in sentences, of concepts and phenomena related to each others, must be able to simultaneously interpret accurately a sequence of multiple semantic representations. A word frequently carries different meanings according to the context. For example, having a ldquobig heartrdquo is usually understood as a quality, whereas the expression ldquobig wordsrdquo has rather a pejorative connotation. The ldquoevaluation polarityrdquo of a word has a probability of changing considerably according to the context. By applying a semantic analysis of co-occurrence expressions, based on the use of phrases having an absolute evaluation polarity, we developed a system achieving analysis capable of: detecting "the role relations between words", the relationship of meaning in a sentence, identifying transitions in the topic, anaphora, endophora, and analyzing even idiomatic expressions and textual emoticons. Our system evaluated correctly ldquopositiverdquo or ldquonegativerdquo nuance for 75.0% of those.
Keywords
computational linguistics; grammars; text analysis; anaphora; context understanding; contextual analysis processing; cooccurrence expression; endophora; evaluation polarity; idiomatic expression; semantic analysis; semantic representation; sentence coherence; sentiments evaluation representation; textual emoticon; word meaning; Dairy products; Data mining; Digital cameras; Heart; Information analysis; Internet; Microwave integrated circuits; Natural languages; Skin; User-generated content; Sentiment analysis; evaluation extraction; semantic coherence;
fLanguage
English
Publisher
ieee
Conference_Titel
Semantic Computing, 2009. ICSC '09. IEEE International Conference on
Conference_Location
Berkeley, CA
Print_ISBN
978-1-4244-4962-0
Electronic_ISBN
978-0-7695-3800-6
Type
conf
DOI
10.1109/ICSC.2009.98
Filename
5298572
Link To Document