To acquire a superior/inferior relation between words corresponding to an application region without causing a cost problem, and to highly precisely retrieve a document suitable for a retrieval input sentence using the relationship.
The document retrieving device is configured to morphologically analyze a retrieval input sentence; to calculate an divergence distance between a word vector of a word B in a morphological analysis result in a word vector database in which a word vector whose component value is the relative value of co-occurrence frequency between an arbitrary word A and a word corresponding to each component or word semantic attributes is associated with the word A, and the word vector in the word vector database of an arbitrary word C; to acquire one or more words C whose distances are small as the related words of the word B; and to acquire a post-replacement morphological analysis result obtained by replacing the word B in the morphological analysis result with the related word of the word B.
UCHIYAMA TOSHIRO
UCHIYAMA MASASHI
JP2002230021A | 2002-08-16 |
JPN7012000834; Thomas Minka: 'Divergence measures and message passing' Microsoft Research Technical Report MSR-TR-2005-173, 20051207
CSNG200700430005; 別所克人 他2名: '単語・意味属性間共起に基づく単語間の階層関係の抽出' 電子情報通信学会技術研究報告 Vol.106 No.518(NLC2006-92), 20070124, 31-36頁, 社団法人電子情報通信学会
Ryuji Ishihara