Retrieval device

Abstract

In systems according to conventional retrieval techniques for restricting a candidate when there are a plurality of candidates as retrieval results, a method involves such a problem that a huge number of candidates are selected in order to prevent omission of candidates, whereby the administration cost is increased, and another method involves such a problem that all candidates are always subjected to processing, whereby the processing period is prolonged and the response performance is lowered A retrieval device according to this invention comprises: a retrieval history storage means which stores a retrieval history including a content inputted from an input means and a candidate list; a narrowing-down system selection means which selects, according to the content of the stored retrieval history, a narrowing-down method from a system of limiting subjects to be retrieved to high-order candidates or a system of re- retrieving on the basis of input in the past and sets retrieval candidates and scores thereof from the retrieval history by the selected narrowing-down system; a candidate score update means which updates the candidate scores by referring to an index for retrieval on the basis of a character string received from the input means; a candidate determination means which determines candidates to be presented, on the basis of the number of the updated candidates and the distribution of the scores; and a candidate presentation means which presents the determined candidates to a user by referring to name information data.

Description

technical field

[0001] The invention relates to a retrieval device for input based on pattern recognition such as text input or input voice, such as names of facilities. Background technique

[0002] The name retrieval technology based on a character string index is a technology for searching a name that appears in a partial character string of a morpheme or an N-character chain of a search object. Japanese Patent No. 3665112 discloses a method of summing up the scores of each candidate based on the collation of partial character strings, and making a high-ranking candidate a search result. In this case, it is possible to realize a fuzzy search that extracts similar candidates even if the character string does not completely match the input. In the fuzzy search, it is necessary to hold a plurality of candidates with different scores, and the memory usage and calculation load are larger than those of the perfect match search.

[0003] The size of the character string index ...

Images