RETRIEVAL DEVICE

• US 20110320464A1
• Filed: 02/09/2010
• Published: 12/29/2011
• Est. Priority Date: 04/06/2009
• Status: Abandoned Application

• Current Assignee: Mitsubishi Electric Corporation
• Original Assignee: Mitsubishi Electric Corporation
• Inventors: Okato, Yohei, Hanazawa, Toshiyuki
• Application Number:
  US13/255,517
• Publication Number:
  US 20110320464A1
• Time in Patent Office: Days
• US Class Current: 707/751
• CPC Class Codes: G06F 16/334 Query execution G06F16/335 takes precedence
• CPC Class Codes: G10L 15/1815 Semantic context, e.g. disambiguation of the recognition hypotheses based on word meaning

Abstract

In conventional retrieval techniques for narrowing down a candidate when there are a plurality of candidates as retrieval results, a method results in a huge number of candidates in order to prevent omission of candidates, whereby the administration cost is increased, and another method involves always processing all candidates, whereby the processing time becomes longer and the response performance is lowered. Retrieval history storage means stores a retrieval history including a content input from input means and a candidate list, narrowing-down method selecting means selects, according to the content of the stored retrieval history, a narrowing-down method from a method of limiting search targets to top-ranked candidates and a method of performing a search again based on inputs made in the past, candidate score update means sets, from the search history, a search candidate and a score thereof according to the selected narrowing-down method, and updates a candidate score based on a character string received from the input means with reference to an index for search, candidate determining means determines a candidate to be presented based on a number of candidates and score distribution which are updated, and candidate presenting means presents, to the user, the determined candidate with reference to name information data.

Claims

1. A retrieval device, comprising:

input means for receiving an input from a user, and outputting a search request;

search history storing means for storing a search history containing a content of the input from the input means and a candidate list;

narrowing-down method selecting means for selecting a narrowing-down method from two methods according to a content of the search history stored in the search history storing means in response to the search request, the two methods comprising a method of limiting search targets to top-ranked candidates and a method of performing a search again based on inputs made in a past;

candidate score updating means for setting, from the search history, a search candidate and a score thereof according to the selected narrowing-down method, and updating a candidate score based on a character string received from the input means with reference to an index for search;

candidate determining means for determining a candidate to be presented based on a number of candidates and a score distribution which are updated by the candidate score updating means; and

candidate presenting means for presenting, to the user, the candidate determined by the candidate determining means with reference to name information data.

2. A retrieval device according to claim 1, further comprising:

a large-vocabulary recognition dictionary for speech recognition; and

means for generating a recognition dictionary for narrowing-down, which generates, when the narrowing-down method selecting means selects the method of limiting search targets to top-ranked candidates, a recognition dictionary for narrowing-down based on name information on target candidates,

wherein the input means receives a voice, and is configured to;

use the recognition dictionary for narrowing-down when the narrowing-down method selecting means selects the method of limiting search targets to top-ranked candidates; and

otherwise, use the large-vocabulary recognition dictionary to perform the speech recognition, and output text.

3. A retrieval device according to claim 2, further comprising means for adapting a recognition dictionary for narrowing-down, which adjusts, when the narrowing-down method selecting means selects the method of performing a search again based on inputs made in a past, the large-vocabulary recognition dictionary based on the search history so that the large-vocabulary recognition dictionary is adapted to an expected narrowing-down speech, and sets the adjusted large-vocabulary recognition dictionary as an adapted recognition dictionary,

wherein the input means receives a voice, and is configured to read the recognition dictionary for narrowing-down or the adapted recognition dictionary according to the narrowing-down method selecting means, to thereby recognize the voice and output text.