Recognition dictionary system and recognition dictionary system updating method

• US 9,020,819 B2
• Filed: 12/28/2006
• Issued: 04/28/2015
• Est. Priority Date: 01/10/2006
• Status: Expired due to Fees

• Current Assignee: Nissan Motor Co., Ltd.
• Original Assignee: Nissan Motor Co., Ltd.
• Inventors: Saitoh, Daisuke, Tomikashi, Minoru, Katsuragawa, Keiko, Oono, Takeshi
• Primary Examiner(s): ROBERTS, SHAUN A
• Application Number:
  US11/646,594
• Publication Number:
  US 20070162281A1
• Time in Patent Office: 3,043 Days
• Field of Search: 704/10, 704/251, 704/270, 704/275
• US Class Current: 704/251
• CPC Class Codes: G06F 40/247 Thesauruses; Synonyms
• CPC Class Codes: G10L 15/197 Probabilistic grammars, e.g. word n-grams

Abstract

A recognition dictionary system includes a variant term accumulation unit, a variant frequency analysis unit, and a recognition dictionary update unit. The variant term accumulation unit is configured to accumulate and store a first variant term that has a meaning equivalent to a base term stored in a recognition dictionary and has first character string information that is different from character string information of the base term. The variant frequency analysis unit is configured to analyze an occurrence frequency of the first variant term accumulated and stored in the variant term accumulation unit and to determine the first variant term as a principal variant term when the occurrence frequency is higher than a first prescribed value. The recognition dictionary update unit is configured to update the recognition dictionary to register the principal variant term in the recognition dictionary as being associated with the base term.

Claims

1. A recognition dictionary system including a client terminal and a server terminal, the recognition dictionary system comprising:

a character string information input unit including

a speech input unit configured and arranged to accept an input of an uttered speech by a user of the client terminal, and

a speech recognition unit configured to extract character string information corresponding to a term included in the uttered speech by the user;

a comprehension control unit configured to determine a response to the user, the comprehension control unit being further configured to

detect whether a first term corresponding to information received by (n)th input by the user is not registered in a recognition dictionary that stores at least character string information of a base term,

output a response that prompts at least (n+1)th input by the user upon detecting that the first term is not registered in the recognition dictionary, and

detect whether a second term corresponding to information received by the (n+1)th input or subsequent input by the user matches the base term stored in the recognition dictionary;

a variant term detection unit configured to detect the first term as a first variant term, which is not registered in the recognition dictionary but that has a meaning equivalent to the base term stored in the recognition dictionary and has character string information that is different from the character string information of the base term, when the second term matches the base term stored in the recognition dictionary, and to associate the first variant term with the base term;

a phoneme recognition unit configured to transform the input of the uttered speech including the first variant term accepted by the speech input unit into first phoneme string information corresponding to the first variant term;

a variant term accumulation unit configured to accumulate and store the first variant term including the first phoneme string information associated with the base term as the first variant term is detected by the variant term detection unit;

a variant frequency analysis unit configured to analyze an occurrence frequency of the first variant term accumulated and stored in the variant term accumulation unit; and

a recognition dictionary update unit configured to update the recognition dictionary to register the first variant term in the recognition dictionary as a variant term of the base term when the occurrence frequency is higher than a first prescribed value that is greater than one.

2. The recognition dictionary system as recited in claim 1, wherein

the client terminal includes the recognition dictionary, and

the server terminal is connected to the client terminal by a communication network, and includes the variant term accumulation unit, the variant frequency analysis unit and the recognition dictionary update unit.

3. The recognition dictionary system as recited in claim 2, wherein

the client terminal includes the character string information input unit,

the server terminal includes

the phoneme recognition unit.

4. The recognition dictionary system as recited in claim 3, wherein

the variant frequency analysis unit of the server terminal is configured to categorize the first variant term by the first phoneme string information based on an inter-phoneme distance with respect to other variant terms of the base term stored in the variant term accumulation unit when analyzing the occurrence frequency of the first variant term.

5. The recognition dictionary system as recited in claim 4, wherein

the variant term detection unit is further configured to detect the first variant term and a second variant term having a meaning equivalent to the base term and including second phoneme string information that is different from phoneme string information of the base term, and to associate the second variant term with the base term,

the recognition dictionary update unit is further configured to update the recognition dictionary to register the second variant term in the recognition dictionary as being associated with the first variant term that has been registered in the recognition dictionary as the variant term of the base term when the first phoneme string information of the first variant term does not perfectly match the second phoneme string information of the second variant term but the first phoneme string information of the first variant term is within a prescribed inter-phoneme distance of the second phoneme string information of the second variant term.

6. The recognition dictionary system as recited in claim 5, wherein

the client terminal is configured to present the first variant term to the user when the uttered speech by the user includes the second variant term.

7. The recognition dictionary system as recited in claim 1, wherein

the comprehension control unit is configured to present a confirmation response to the information received by the (n)th input by the user, and to detect that the first term is not registered in the recognition dictionary if the information received by the (n+1)th input by the user in response to the confirmation response includes at least one of a correction expression and a negative expression.

8. The recognition dictionary system as recited in claim 7, wherein

the client terminal further includes

a speech input unit configured and arranged to accept an input of an uttered speech by the user, and

a speech recognition unit configured to recognize a recognition term candidate when the uttered speech has a prescribed recognition score, and

the comprehension control unit is configured to detect the first term is not registered in the recognition dictionary when the recognition term candidate is not obtained by the uttered speech in the (n)th input by the user.

9. The recognition dictionary system as recited in claim 8, wherein

the comprehension control unit is configured to output a verification response to confirm with the user that the first term is a variant term of the base term upon detecting that the second term matches the base term stored in the recognition dictionary, and

the variant term detection unit is configured to determine the first term as the first variant term when a response by the user to the verification response indicates that the user positively acknowledges that the first term is the variant term of the base term.

10. The recognition dictionary system as recited in claim 9, wherein

the comprehension control unit is configured to output a response that prompts the at least (n+1)th input other than the input of the uttered speech upon detecting that the first term is not registered in the recognition dictionary.

11. The recognition dictionary system as recited in claim 2, wherein

the variant frequency analysis unit is configured to refer to the variant term accumulation unit at a prescribed time interval, the variant frequency analysis unit being further configured to determine that the first variant term is a deletion target variant term when the occurrence frequency of the variant term falls below a second prescribed value that is smaller than the first prescribed value, and

the recognition dictionary update unit is further configured to delete the first variant term from the recognition dictionary when the variant frequency analysis unit determines that the first variant term is the deletion target variant term.

12. The recognition dictionary system as recited in claim 2, wherein

the client terminal is configured to accept information corresponding to the first variant term that is input by a user for setting a destination with a navigation system.

13. The recognition dictionary system as recited in claim 12, wherein

the variant term accumulation unit is configured to accumulate and store user attribute information including at least one of a residential area, an age bracket, a family composition, and a vehicle attribute of the user with the user attribute information being associated with the base term,

the variant frequency analysis unit is configured to analyze the occurrence frequency of the first variant term for each user attribute information, and

the recognition dictionary update unit is configured to register the first variant term as the variant term of the base term when the occurrence frequency is higher than the first prescribed value in the recognition dictionary used by the user to whom the user attribute information corresponding to the first variant term belongs.

14. A recognition dictionary system including a client terminal and a server terminal, the recognition dictionary system comprising:

character string information inputting means for accepting an input of an uttered speech by a user of the client terminal and for extracting character string information corresponding to a term included in the uttered speech by the user;

comprehension controlling means for determining a response to the user, and including functions for

detecting whether a first term corresponding to information received by (n)th input by the user is not registered in a recognition dictionary that stores at least character string information of a base term,

outputting a response that prompts at least (n+1)th input by the user upon detecting that the first term is not registered in the recognition dictionary, and

detecting whether a second term corresponding to information received by the (n+1)th input or subsequent input by the user matches the base term stored in the recognition dictionary;

variant term detecting means for detecting the first term as a first variant term, which is not registered in the recognition dictionary but has a meaning equivalent to the base term and has different character string information from the character string information of the base term, when the second term matches the base term stored in the recognition dictionary, and for associating the first variant term with the base term;

phoneme recognition means for transforming the input of the uttered speech including the first variant term accepted by the speech input unit into first phoneme string information corresponding to the first variant term;

variant term accumulating means for accumulating and storing the first variant term including the first phoneme string information associated with the base term as the first variant term is detected;

frequency occurrence analyzing means for analyzing an occurrence frequency of the first variant term accumulated and stored by the variant term accumulating means; and

dictionary updating means for updating the recognition dictionary to register the first variant term in the recognition dictionary as a variant term of the base term when the occurrence frequency is higher than a first prescribed value that is greater than one.

15. A recognition dictionary system updating method for updating a recognition dictionary system including a client terminal and a server terminal, the recognition dictionary system updating method comprising:

accepting an input of an uttered speech by a user of the client terminal;

extracting character string information corresponding to a term included in the uttered speech by the user;

detecting whether a first term corresponding to information received by (n)th input by the user is not registered in a recognition dictionary that stores at least character string information of a base term;

outputting a response that prompts at least (n+1)th input by the user upon detecting that the first term is not registered in the recognition dictionary;

detecting whether a second term corresponding to information received by the (n+1)th input or subsequent input by the user matches the base term stored in the recognition dictionary;

detecting the first term as a first variant term, which is not registered in the recognition dictionary but has a meaning equivalent to the base term and has different character string information from the character string information of the base term, when the second term matches the base term stored in the recognition dictionary, and associating the first variant term with the base term;

transforming the input of the uttered speech including the first variant term accepted by the speech input unit into first phoneme string information corresponding to the first variant term;

accumulating and storing the first variant term including the first phoneme string information associated with the base term as the first variant term is detected;

analyzing an occurrence frequency of the first variant term accumulated and stored; and

updating the recognition dictionary to register the first variant term in the recognition dictionary as a variant term of the base term when the occurrence frequency is higher than a first prescribed value that is greater than one.