SNIPPET GENERATION DEVICE, SNIPPET GENERATION METHOD AND SNIPPET GENERATION PROGRAM

摘要

PROBLEM TO BE SOLVED: To properly generate a snippet from a structured document on the basis of a retrieval query.;SOLUTION: A snippet generation device 1 is configured so that: a DOM tree construction unit 11 performs syntactic analysis of a structured document, for expanding respective nodes forming the document into a tree structure, for extracting nodes of a title and a content of the document from the structure. A cluster generation unit 12 performs clustering of the respective nodes, based on similarity of respective nodes, of the tree structure. A score application unit 13 applies a score to the respective cluster, based on a word of a retrieval query, a related word, and a unique expression of the word, the retrieval query is included in the cluster generated by the clustering. a snippet generation unit 14 selects clusters having top rank score, in which length of the generated snippet is equal to or less than a threshold, as candidates of an element of the snippet, and arranges again the selected clusters in an order of appearance of the structured document, for generating as the snippet.;SELECTED DRAWING: Figure 1;COPYRIGHT: (C)2016,JPO&INPIT