RECOGNITION DICTIONARY CREATION DEVICE AND VOICE RECOGNITION DEVICE

• US 20120226491A1
• Filed: 02/05/2010
• Published: 09/06/2012
• Est. Priority Date: 02/05/2010
• Status: Active Grant

• Current Assignee: Mitsubishi Electric Corporation
• Original Assignee: Mitsubishi Electric Corporation
• Inventors: Yamazaki, Michihiro, Ishii, Jun, Ishikawa, Yasushi
• Granted Patent: US 8,868,431 B2
• Time in Patent Office: Days
• US Class Current: 704/8
• CPC Class Codes: G10L 15/06 Creation of reference templates; Training of speech recognition systems, e.g. adaptation to the characteristics of the speaker's voice G10L15/14 takes precedence
• CPC Class Codes: G10L 2015/025 Phonemes, fenemes or fenones being the recognition units

Abstract

A recognition dictionary creation device identifies the language of a reading of an inputted text which is a target to be registered and adds a reading with phonemes in the language identified thereby to the target text to be registered, and also converts the reading of the target text to be registered from the phonemes in the language identified thereby to phonemes in a language to be recognized which is handled in voice recognition to create a recognition dictionary in which the converted reading of the target text to be registered is registered.

Claims

1. A recognition dictionary creation device comprising:

a language identification unit for identifying a language of a reading of an inputted text which is a target to be registered;

a reading addition unit for adding a reading with phonemes in the language identified by said language identification unit to said target text to be registered;

a reading conversion unit for converting said reading of said target text to be registered from the phonemes in the language identified by said language identification unit to phonemes in a language to be recognized which is handled in voice recognition; and

a recognition dictionary creation unit for creating a recognition dictionary in which said reading of said target text to be registered which is converted by said reading conversion unit is registered.

2. The recognition dictionary creation device according to claim 1, wherein said language identification unit outputs a predetermined number of languages each having a score showing a higher likelihood that the reading of said target text to be registered is described in the language, among a plurality of languages which are targets for language identification, as results of the identification, said reading addition unit adds a reading with phonemes in each of said predetermined number of languages identified by said language identification unit to said target text to be registered, and said reading conversion unit converts the reading of said target text to be registered from the phonemes in each of said predetermined number of languages identified by said language identification unit to phonemes in the language to be recognized.

3. The recognition dictionary creation device according to claim 2, wherein said language identification unit outputs said language to be recognized as a result of the identification when said score is smaller than a predetermined threshold.

4. The recognition dictionary creation device according to claim 1, wherein said recognition dictionary creation device includes an object-to-be-excluded storage unit for storing a character or a character string which is to be excluded from a target for language identification, a language identification pre-processing unit for excluding a portion corresponding to said character or said character string which is to be excluded from said target for language identification and which is stored in said object-to-be-excluded storage unit from said target text to be registered, and a process descriptions storage unit for storing descriptions of the excluding process of excluding said character or said character string which is to be excluded from said target for language identification which is performed on said target text to be registered by said language identification pre-processing unit, and wherein said language identification unit identifies the language of the reading of said target text to be registered from which said character or said character string which is to be excluded from said target for language identification is excluded by said language identification pre-processing unit, and said recognition dictionary creation unit refers to the descriptions of the excluding process stored in said process descriptions storage unit to determine the reading of said target text to be registered from a reading of said character or said character string which is to be excluded from said target for language identification, and a reading of said target text to be registered from which said character or said character string which is to be excluded from said target for language identification is excluded.

5. The recognition dictionary creation device according to claim 4, wherein said recognition dictionary creation device includes a dividing object storage unit for storing a dividing character or a dividing character string, said language identification pre-processing unit divides a character string of said target text to be registered into divisions by using said dividing character or said dividing character string stored in said dividing object storage unit, descriptions of the dividing process which is performed on said target text to be registered by said language identification pre-processing unit are stored in said process descriptions storage unit, said language identification unit identifies a language of a reading of each of the divisions into which said target text to be registered is divided by said language identification pre-processing unit, and said recognition dictionary creation unit refers to the descriptions of the dividing process stored in said process descriptions storage unit to determine the reading of said target text to be registered from the reading of each of said divisions and create a recognition dictionary in which this reading of said target text to be registered is registered.

6. The recognition dictionary creation device according to claim 5, wherein said dividing object storage unit stores a dividing character or a dividing character string including a numeral or an uppercase character string, and a reading of the dividing character or the dividing character string in the language to be recognized, said language identification pre-processing unit excludes the dividing character or the dividing character string including said numeral or said uppercase character string from said target text to be registered, and divides the character string of said target text to be registered into divisions, said language identification unit identifies the language of the reading of each of the divisions into which said target text to be registered is divided by said language identification pre-processing unit, said reading addition unit adds a reading in a language of said divisions which is identified by said language identification unit to the dividing character or the dividing character string including said numeral or said uppercase character string, and said recognition dictionary creation unit refers to the descriptions of the excluding process stored in said process descriptions storage unit to determine the reading of said target text to be registered from the reading of each of said divisions, a reading in the language to be recognized of said dividing character or said dividing character string, and a reading of each of said divisions, into which said target text to be registered is divided, in said language to be recognized into which the reading in the language of said each of said divisions is converted, and create a recognition dictionary in which said reading of said target text to be registered is registered.

7. The recognition dictionary creation device according to claim 1, wherein said language identification unit uses a language identification model including an N-gram for each language which is a target for language identification, and its probability of occurrence to calculate a sequence probability of an N-gram of said target text to be registered for each language, and identify the language of the reading from a likelihood based on said sequence probability.

8. The recognition dictionary creation device according to claim 7, wherein said language identification unit replaces each character or each character string which is included in said target text to be registered and is described in common in a plurality of languages which are targets for the language identification and which does not contribute to the language identification with a special character to create an N-gram.

9. The recognition dictionary creation device according to claim 7, wherein when a character or a character string which is used for a limited number of languages is included in said target text to be registered, said language identification unit outputs a language having a highest likelihood, among said limited number of languages, as a result of the identification.

10. The recognition dictionary creation device according to claim 7, wherein said language identification unit is provided with a language identification model for each process using a word to be recognized, and uses a language identification model corresponding to a process using said target text to be registered for the language identification.

11. A voice recognition device comprising:

a language identification unit for identifying a language of a reading of an inputted text which is a target to be registered;

a reading addition unit for adding a reading with phonemes in the language identified by said language identification unit to said target text to be registered;

a reading conversion unit for converting said reading of said target text to be registered from the phonemes in the language identified by said language identification unit to phonemes in a language to be recognized which is handled invoice recognition;

a recognition dictionary creation unit for creating a recognition dictionary in which said reading of said target text to be registered which is converted by said reading conversion unit is registered; and

a voice recognition unit for performing voice recognition on a voice inputted thereto by making reference to said recognition dictionary created by said recognition dictionary creation unit.

12. A voice recognition device comprising:

an object-to-be-excluded storage unit for storing a character or a character string which is to be excluded from a target for language identification;

a dividing object storage unit for storing a dividing character or a dividing character string;

a language identification pre-processing unit for excluding said character or said character string which is to be excluded from said target for language identification from an inputted target text to be registered and also dividing said inputted target text to be registered into divisions by using said dividing character or said dividing character string on a basis of the contents stored in said object-to-be-excluded storage unit and the contents stored in said dividing object storage unit;

a process descriptions storage unit for storing descriptions of the process which is performed on said target text to be registered by said language identification pre-processing unit;

a language identification unit for identifying a language of a reading of said target text to be registered on which said process is performed by said language identification pre-processing unit;

a reading addition unit for adding a reading with phonemes in the language identified by said language identification unit to said target text to be registered;

a reading conversion unit for converting the reading of said target text to be registered from the phonemes in the language identified by said language identification unit to phonemes in a language to be recognized which is handled in voice recognition;

a recognition dictionary creation unit for referring to the descriptions of said process stored in said process descriptions storage unit to determine a reading of said target text to be registered on which said process is performed by said language identification pre-processing unit and create a recognition dictionary in which this reading is registered; and

a voice recognition unit for performing voice recognition on a voice inputted thereto by making reference to said recognition dictionary created by said recognition dictionary creation unit.