Recognizing device, computer-readable recording medium, recognizing method, generating device, and generating method

• US 9,082,404 B2
• Filed: 08/15/2012
• Issued: 07/14/2015
• Est. Priority Date: 10/12/2011
• Status: Active Grant

• Current Assignee: Fujitsu Limited
• Original Assignee: Fujitsu Limited
• Inventors: Harada, Shouji
• Primary Examiner(s): Desir, Pierre-Louis
• Assistant Examiner(s): THOMAS-HOMESCU, ANNE L
• Application Number:
  US13/586,533
• Publication Number:
  US 20130096918A1
• Time in Patent Office: 1,063 Days
• Field of Search: 704/251, 704/249, 704/254, 704/9, 704/257, 704/231, 704/235, 704/275, 704/266, 704/243, 704/252, 704/270, 704/2, 704/246, 704/236, 704/10, 704/255, 704/270.1, 704/239, 704/240, 704/8
• US Class Current: 1/1
• CPC Class Codes: G10L 15/10 using distance or distortion measures between unknown speech and reference templates
• CPC Class Codes: G10L 15/19 Grammatical context, e.g. disambiguation of the recognition hypotheses based on word sequence rules
• CPC Class Codes: G10L 2015/027 Syllables being the recognition units

Abstract

A recognizing device includes a memory and a processor coupled to the memory. The memory stores words included in a text and positional information indicating a position of the words in the text. The processor executes a process including comparing an input voice signal with reading information of a character string that connects a plurality of words stored in the memory to calculate a similarity; calculating a connection score indicating a proximity between the plurality of connected words based on positional information of the words stored in the memory; and determining a character string corresponding to the voice signal based on the similarity and the connection score.

Claims

1. A recognizing device, comprising:

a memory that stores words included in a text and positional information indicating a position of the words in an ascending numerical order from a heading in the text; and

a processor coupled to the memory, wherein the processor executes a process comprising;

comparing an input voice signal with reading information of a character string that connects a plurality of words stored in the memory unit to calculate a similarity;

calculating a connection score indicating a proximity between the plurality of connected words based on positional information indicating a position of the words in an ascending numerical order from a heading in the text, wherein the words are stored in the memory unit;

determining a character string corresponding to the voice signal based on a value obtained by adding the similarity and the connection score which exceeds a threshold value, thereby recognizing connected words which are included in the sentence but not adjacent to each other in the sentence and generating a dictionary that allows precise recognition of voice.

2. The recognizing device according to claim 1, wherein the memory further stores a word string in which a plurality of words are coupled and positional information indicating the position of the word string in the text,

the comparing includes calculating the similarity using the word string as a word, and

the calculating the connection score includes calculating the connection score using the word string as a word.

3. The recognizing device according to claim 1, wherein the memory further stores identification information for identifying a sentence so as to be associated with the sentence included in the text, and

the calculating the connection score includes calculating the connection score indicating the proximity between the plurality of connected words for every sentence based on the positional information and the identification information of the words stored in the memory.

4. The recognizing device according to claim 2, wherein the memory further stores identification information for identifying a sentence so as to be associated with the sentence included in the text, and

the calculating the connection score includes calculating the connection score indicating the proximity between the plurality of connected words for every sentence based on the positional information and the identification information of the words stored in the memory.

5. The recognizing device according to claim 1, wherein the memory further stores information indicating a probability that a second word is connected to a first word, and

the calculating the connection score includes further correcting the connection score based on the information indicating the probability that the second word is connected to the first word stored in the memory.

6. The recognizing device according to claim 2, wherein the memory further stores information indicating a probability that a second word is connected to a first word, and

the calculating the connection score includes further correcting the connection score based on the information indicating the probability that the second word is connected to the first word stored in the memory.

7. The recognizing device according to claim 3, wherein the memory further stores information indicating a probability that a second word is connected to a first word, and

the calculating the connection score includes further correcting the connection score based on the information indicating the probability that the second word is connected to the first word stored in the memory.

8. The recognizing device according to claim 4, wherein the memory further stores information indicating a probability that a second word is connected to a first word, and

the calculating the connection score includes further correcting the connection score based on the information indicating the probability that the second word is connected to the first word stored in the memory.

9. The recognizing device according to claim 1, wherein the memory further stores information concerning the suitability of the combination of word classes of the plurality of connected words, and

the calculating the connection score includes further correcting the connection score based on the information concerning the suitability of the combinations of word classes of the plurality of connected words stored in the memory.

10. A recognizing method that is executed by a computer, the recognizing method comprising:

storing in memory words included in a text and positional information indicating a position of the words in an ascending numerical order from a heading in the text;

comparing, using a processor, an input voice signal with reading information of a character string that connects a plurality of the words included in a text and positional information indicating a position of the words in an ascending numerical order from a heading in the text to calculate a similarity;

calculating a connection score indicating a proximity between the plurality of connected words based on positional information indicating a position of the words in an ascending numerical order from a heading in the text, wherein the words are stored in the memory unit; and

determining a character string corresponding to the voice signal based on a value obtained by adding the similarity and the connection score which exceeds a threshold value, thereby recognizing connected words which are included in the sentence but not adjacent to each other in the sentence and generating a dictionary that allows precise recognition of voice.