C-2-1 感情モデルを用いた対話システムの提案(評価・調査分析(2))

概要

論文の詳細を見る
In recent years, development of interactive artificial intelligence is remarkable, but there are few models that can be human-like conversation yet. Emotional element is mentioned as one of the causes. To prevent this problem, discovery and analysis of emotion contained in the document is a very important issue. There is Plutchik emotion model that as one method to analyze the feelings. That sees emotions as consists of a combination of eight primary emotions. In this study, we try to introduce Plutchik emotion model into the conversation system. For emotion analysis of text using an emotion model, the emotion value must be given to the words that are the smallest unit of text. But, it put a lot of labor to obtain it with set out a questionnaire. In this study, we constructed the lexical network from a sentence of interpretation of a word, for the purpose of determining the emotion values about a lot more words from the basis words. As a result, we obtained lexical network with the nodes of about 500,000 that consist of two links of the same orientation link and the reverse orientation link.
2012-12-26