| 摘要 |
<P>PROBLEM TO BE SOLVED: To provide an emotion estimation device, method, program, and recording medium, to estimate emotions from target voice data or target text data without using a database, for example, an emotion related word DB, which stores words' semantic information. <P>SOLUTION: Explicitness feature quantities expressing to what extent the semantic content is explicitly indicated in target text data is estimated by using at least one of a morphological analysis result or a syntax analysis result of the target text data. Whether or not an emotion of anger is expressed in the target text data is identified from the explicitness feature quantities of the target text data estimated by the explicitness estimation part by using a discriminator preliminarily learned from the explicitness feature quantities from text data for learning including a teacher signal showing whether or not an emotion of anger is expressed. <P>COPYRIGHT: (C)2012,JPO&INPIT |
