Speech recognition device and method outputting or rejecting derived words

• US 8,903,724 B2
• Filed: 02/01/2012
• Issued: 12/02/2014
• Est. Priority Date: 03/15/2011
• Status: Active Grant

• Current Assignee: Fujitsu Limited
• Original Assignee: Fujitsu Limited
• Inventors: Washio, Nobuyuki, Harada, Shouji
• Primary Examiner(s): Lerner, Martin
• Application Number:
  US13/363,411
• Publication Number:
  US 20120239402A1
• Time in Patent Office: 1,035 Days
• Field of Search: 704/231, 704/236, 704/251, 704/252, 704/257, 707/765, 707/766, 707/767
• US Class Current: 704/252
• CPC Class Codes: G10L 15/02 Feature extraction for speech recognition; Selection of recognition unit
• CPC Class Codes: G10L 2015/088 Word spotting

Abstract

A speech recognition device includes, a speech recognition section that conducts a search, by speech recognition, on audio data stored in a first memory section to extract word-spoken portions where plural words transferred are each spoken and, of the word-spoken portions extracted, rejects the word-spoken portion for the word designated as a rejecting object; an acquisition section that obtains a derived word of a designated search target word, the derived word being generated in accordance with a derived word generation rule stored in a second memory section or read out from the second memory section; a transfer section that transfers the derived word and the search target word to the speech recognition section, the derived word being set to the outputting object or the rejecting object by the acquisition section; and an output section that outputs the word-spoken portion extracted and not rejected in the search.

Claims

1. A speech recognition device comprising:

a processor; and

a memory which stores a plurality of instructions, which when executed by the processor, cause the processor to execute,

conducting a search, by speech recognition, on audio data stored in a first memory to extract wordspoken portions where plural words transferred are each spoken and, of the word-spoken portions extracted, rejects the word-spoken portion for the word designated as a rejecting object;

obtaining a derived word of a designated search target word, the derived word being generated in accordance with a derived word generation rule stored in a second memory or read out from the second memory where the derived word is stored in association with the search target word, and sets the derived word to an outputting object or the rejecting object according to setting designation information or a user'"'"'s instruction, the setting designation information being stored in the second memory and indicating whether the derived word is the outputting object or the rejecting object;

transferring the derived word and the search target word to the conducting, the derived word being obtained and set to the outputting object or the rejecting object by the obtaining; and

outputting the word-spoken portion extracted and not rejected in the search of the conducting,

wherein the derived word generation rule is generated based on a difference between the search target word and a derived word that is stored in the second memory.

2. The device according to claim 1,

wherein the second memory stores the setting designation information for each of the derived word generation rules or for each of the derived words stored in the second memory, and

wherein the processor sets the derived word obtained to the outputting object or the rejecting object based on the setting designation information that is stored in the second memory and corresponds to the derived word.

3. The device according to claim 1,

wherein the second memory stores the setting designation information by type of word, the type being set for a word that is to be designated as the search target word, and

wherein the processor sets the derived word to the outputting object or the rejecting object based on the setting designation information that is stored in the second memory and corresponds to the type of the search target word designated.

4. The device according to claim 1,

wherein the processor presents the derived word, and sets the derived word presented to the outputting object or the rejecting object based on an instruction.

5. The device according to claim 1, further comprising:

presenting to a user the derived word and a result of the setting to the outputting object or the rejecting object, which is made by the processor with respect to the derived word, and changing the setting of the outputting object or the rejecting object with respect to the derived word as to which the user instructed changing the setting of the outputting object or the rejecting object.

6. The device according to claim 5,

wherein the second memory stores the setting designation information by the derived word generation rule or by the derived word stored in the second memory, and

wherein the processor changes the setting designation information in response to the user'"'"'s instruction, the setting designation information being stored in the second memory and corresponding to the derived word as to which the user instructed changing the setting of the outputting object or the rejecting object.

7. The device according to claim 1,

wherein the second memory stores the derived word generation rule by type of word and type identifying information, the type being set for a word that is to be designated as the search target word, the type identifying information identifying the type of a word that is to be designated as the search target word, and

wherein the processor identifies the type of the search target word based on the type identifying information stored in the second memory, and generates the derived word of the search target word in accordance with the derived word generation rule that corresponds to the type of the search target word identified.

8. The device according to claim 7,

wherein the second memory stores the derived words for a subset of words that are to be designated as the search target words, in association with respective words of the subset,

wherein the processor obtains the derived word of the search target word by

reading out the derived word stored in the second memory in association with the search target word when the derived word is stored in the second memory in association with the search target word, and

generating the derived word of the search target word in accordance with the derived word generation rule corresponding to the type of the search target word after identifying the type of the search target word when the derived word is not stored in the second memory in association with the search target word.

9. The device according to claim 1,

wherein the processor is configured to

calculate an evaluation value for evaluating a degree of matching between a pronunciation of a word for evaluation and audio sound in a portion to be evaluated, the portion being included in audio sound represented by the audio data, and

determine whether or not the portion to be evaluated is a portion where the word for evaluation is spoken based on whether or not the evaluation value calculated is equal to or larger than a threshold value, and

wherein the threshold value in a case where the derived word is used as the word for evaluation is set to less than that in a case where the search target word is used as the word for evaluation.

10. The device according to claim 1,

wherein the derived word is a word generated by adding an affix to the search target word, a word generated by removing an affix included in the search target word, a synonym of the search target word, a word having a same meaning but in a different language, a word being different in number from the search target word, or a word in a different tense from the search target word.

11. A speech recognition method comprising:

obtaining a derived word of a designated search target word, the derived word being generated in accordance with a derived word generation rule stored in a second memory section or read out from the second memory section where the derived word is stored in association with the search target word;

setting, by a processor, the derived word by the obtaining to an outputting object or a rejecting object according to setting designation information or a user'"'"'s instruction, the setting designation information being stored in the second memory section and indicating whether the derived word is the outputting object or the rejecting object;

searching, by speech recognition, audio data stored in a first memory section to extract word-spoken portions where the search target word and the derived word are each spoken, the derived word being obtained by the obtaining and set to the outputting object or the rejecting object by the setting;

rejecting, of the word-spoken portions extracted by the searching, the word-spoken portion for the derived word that has been set to the rejecting object, and

outputting the word-spoken portion extracted by the searching and not rejected by the rejecting,

wherein the derived word generation rule is generated based on a difference between the search target word and a derived word that is stored in the second memory.

12. The method according to claim 11,

wherein the second memory section stores the setting designation information by the derived word generation rule or by the derived word stored in the second memory section, and

wherein the obtaining sets the derived word to the outputting object or the rejecting object based on the setting designation information that is stored in the second memory section and corresponds to the derived word by the obtaining.

13. The method according to claim 11,

wherein the second memory section stores the setting designation information by type of word, the type being set for a word that is to be designated as the search target word, and

wherein the obtaining further sets the derived word to the outputting object or the rejecting object based on the setting designation information that is stored in the second memory section and corresponds to the type of the search target word designated.

14. The method according to claim 11,

wherein the obtaining presents the derived word by the obtaining, and sets the derived word presented to the outputting object or the rejecting object based on an instruction.

15. The method according to claim 11, further comprising:

changing the setting of the outputting object or the rejecting object with respect to the derived word as to which a user instructed to change the setting of the outputting object or the rejecting object after presenting to the user the derived word by the obtaining and a result of the setting to the outputting object or the rejecting object, the setting being made with respect to the derived word by the obtaining.

16. The method according to claim 15,

wherein the second memory section stores the setting designation information by the derived word generation rule or by the derived word stored in the second memory section, and

wherein the changing changes the setting designation information in response to the user'"'"'s instruction, the setting designation information being stored in the second memory section and corresponding to the derived word as to which the user instructed to change the setting of the outputting object or the rejecting object.

17. The method according to claim 11,

wherein the second memory section stores the derived word generation rule by type of word and type identifying information, the type being set for a word that is to be designated as the search target word, the type identifying information identifying the type of a word that is to be designated as the search target word, and

wherein the obtaining further identifies the type of the search target word based on the type identifying information stored in the second memory section, and generates the derived word of the search target word in accordance with the derived word generation rule that corresponds to the type of the search target word identified.

18. The method according to claim 17,

wherein the second memory section stores the derived words for a subset of words that are to be designated as the search target words, in association with respective words of the subset,

wherein the obtaining further obtains the derived word of the search target word by

reading out the derived word stored in the second memory section in association with the search target word when the derived word is stored in the second memory section in association with the search target word, and

generating the derived word of the search target word in accordance with the derived word generation rule corresponding to the type of the search target word after identifying the type of the search target word when the derived word is not stored in the second memory section in association with the search target word.

19. The method according to claim 11,

wherein the derived word is a word generated by adding an affix to the search target word, a word generated by removing an affix included in the search target word, a synonym of the search target word, a word having a same meaning but in a different language, a word being different in number from the search target word, or a word in a different tense from the search target word.

20. A non-transitory computer-readable storage medium storing a speech recognition program that causes a computer to execute a process comprising:

obtaining a derived word of a designated search target word, the derived word being generated in accordance with a derived word generation rule stored in a second memory or read out from the second memory where the derived word is stored in association with the search target word;

setting the derived word by the obtaining to either an outputting object or a rejecting object according to setting designation information or a user'"'"'s instruction, the setting designation information being stored in the second memory and indicating whether the derived word is the outputting object or the rejecting object;

searching, by speech recognition, audio data stored in a first memory to extract word-spoken portions where the search target word and the derived word are each spoken, the derived word being obtained by the obtaining and set to either the outputting object or the rejecting object by the setting;

rejecting, of the word-spoken portions extracted by the searching, the word-spoken portion for the derived word that has been set to the rejecting object; and

outputting the word-spoken portion extracted by the searching and not rejected by the rejecting,

wherein the derived word generation rule is generated based on a difference between the search target word and a derived word that is stored in the second memory.