摘要 |
An information processor carrying out statistical natural language processing for a document, the information processor includes a characteristic amount extraction unit configured to detect context information including a proper noun pair from the document and extract a characteristic amount of the detected context information; a characteristic amount analysis unit configured to, by analyzing the characteristic amount of the extracted context information using a probability model in which a document topic meaning an entire topic of the document and a context topic meaning a local topic of the document are considered, estimate a potential variable and a context topic ratio in the probability model; and a clustering unit configured to cluster the proper noun pair included in the context information based on the context topic ratio estimated regarding the characteristic amount of the respective context information. |