| 摘要 |
<p><P>PROBLEM TO BE SOLVED: To establish statistical natural language processing using context information under the consideration of possibility that such assumption that "neighboring words are mutually related" is not established even on the basis of such assumption. <P>SOLUTION: An information processing apparatus includes: a step S1 that a processing object document is input to a characteristic amount extraction part; a step S2 that the characteristic amount extraction part extracts characteristic amounts for each context information included in the processing object document; a step S3 that a characteristic amount analysis part estimates latent variables corresponding to the characteristic amounts of each context information of the processing object document by gypsum sampling; a step S4 that a clustering processing part considers the context topic ratio of each context as a new characteristic amount vector, and performs the clustering of (the unique noun pair of) the context information based on the characteristic amount vector; and a step S5 that a basic information generation part generates basic information based on the clustering result stored in the analysis result DB. This invention may be applied to the statistical natural language processing of the document. <P>COPYRIGHT: (C)2011,JPO&INPIT</p> |
