Systems and methods for word prediction and speech recognition

• US 20020038207A1
• Filed: 07/11/2001
• Published: 03/28/2002
• Est. Priority Date: 07/11/2000
• Status: Active Grant

• Current Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Original Assignee: International Business Machines Corporation
• Inventors: Mori, Shinsuke, Nishimura, Masafumi, Itoh, Nobuyasu
• Granted Patent: US 7,359,852 B2
• Time in Patent Office: Days
• US Class Current: 704/9
• CPC Class Codes: G10L 15/19 Grammatical context, e.g. disambiguation of the recognition hypotheses based on word sequence rules

Abstract

A word prediction method that improves the precision accuracy, and a speech recognition method and an apparatus therefor are provided. For the prediction of a sixth word “?”, a partial analysis tree having a modification relationship with the sixth word is predicted. “sara-ni sho-senkyoku no” has two partial analysis trees, “sara-ni” and “sho-senkyoku no”. It is predicted that “sara-ni” does not have a modification relationship with the sixth word, and that “sho-senkyoku no” does. Then, “donyu”, which is the sixth word from “sho-senkyoku no”, is predicted. In this example, since “sara-ni” is not useful information for the prediction of “donyu”, it is preferable that “donyu” be predicted only by “sho-senkyoku no”.

Claims

1. A word prediction method, said method comprising the steps of:

specifying a sentence structure consisting of multiple words, including a target word to be predicted;

employing said sentence structure to select a word and/or a word sequence that has a modification relationship with said target word to be predicted; and

predicting said target word based on said word and/or said word sequence that is selected.

2. The word prediction method according to claim 1, wherein said word and/or said word sequence constitute a partial analysis tree structure in said sentence structure.

3. The word prediction method according to claim 1, wherein, when multiple words and/or word sequences are selected, word prediction is performed based on said words and/or word sequences that are selected.

4. A word prediction method, said method comprising the steps of:

specifying a modification of a word to be predicted by a different word and/or word sequence; and

prediction said word using said different word and/or word sequence that is specified.

5. The word prediction method according to claim 4, wherein said modification includes a modification direction, and said word to be predicted modifies a prior word.

6. The word prediction method according to claim 4, wherein, when multiple modifications are established between said word to be predicted and said different word and/or word sequence, a word is predicted for each of said modifications.

7. A speech recognition method, said method comprising the steps of:

specifying the structure of a phrase from the beginning of a sentence to the j-th word (j=0, 1, 2, . . . );

employing a sentence structure up to said j-th word to specify one or multiple partial analysis trees modifying the (j+1)th word;

predicting said (j+1)th word based on said one or multiple partial analysis trees;

obtaining a sentence for a phrase including said predicted (j+1)th word and the probability value for said sentence structure; and

performing the above steps up to the last word of said sentence, and selecting, as speech recognition results, a sentence structure and a word sequence having maximum probability values.

8. A word prediction apparatus comprising:

a storer which stores a dictionary wherein text data that have been learned is written; and

a predictor which employs a structure of a sentence consisting of multiple words, including a word to be predicted, to select a word and/or a word sequence that has a modification relationship with said word to be predicted, and which predicts said word based on said word and/or said word sequence that is selected.

9. A word prediction apparatus comprising:

a storer which stores a dictionary wherein text data that have been learned is written; and

a predictor which selects a word and/or a word sequence that has a modification relationship with a word to be predicted, and which predicts said word based on said word and/or said word sequence that is selected.

10. The word prediction apparatus according to claim 9, wherein, for a predetermined word or word sequence based on said text data, a frequency where at another predetermined word appears is stored in said dictionary.

11. The word prediction apparatus according to claim 10, wherein said predictor calculates a probability value for a word sequence including a word that is predicted based on said frequency.

12. The word prediction apparatus according to claim 11, wherein said predictor selects, as a prediction result, a word sequence having the maximum probability value.

13. A speech recognition apparatus comprising:

an acoustic processor which converts an input analog speech signal into a digital signal;

a first storer which stores an acoustic model that has learned a feature of speech;

a second storer which stores a dictionary wherein an appearance frequency of a predetermined word relative to another predetermined word and/or word sequence is written; and

a recognizer which uses said acoustic model and said dictionary to calculate a probability value for said digital signal, and which recognizes a word having the maximum probability value as input speech, wherein said recognizer predicts a word to be predicted based on a structure of a sentence including said word, and employs said appearance frequency to calculate said probability value for said sentence, including said word that is predicted.

14. A computer system comprising:

a specifier which specifies a modification relationship between a word to be predicted and another word and/or word sequence, and which predicts said word by employing said word and/or word sequence modifying said word; and

a display which displays said word that is predicted, and said modification relationship.

15. A storage medium, on which a computer readable program is stored that permits a computer to perform:

a first process for specifying a sentence structure consisting of multiple words, including a target word to be predicted;

a second process for employing said sentence structure to select a word and/or a word sequence that has a modification relationship with said target word to be predicted; and

a third process for predicting said target word based on said word and/or said word sequence that is selected.

16. A storage medium, on which a computer readable program is stored, that permits a computer to perform:

a first process for specifying a modification relationship between a word to be predicted and a different word and/or word sequence; and

a second process for predicting said word using said different word and/or word sequence that is specified.

17. A program transmission apparatus comprising:

a storer which stores a program permitting a computer to perform

a first process for specifying a sentence structure of a multiple word sentence, including a target word to be predicted,

a second process for employing said sentence structure to select a word and/or a word sequence that has a modification relationship with said target word to be predicted, and

a third process for predicting said target word based on said word and/or said word sequence that is selected; and

a transmitter which reads said program from said storer and transmits said program.

18. A program transmission apparatus comprising:

a storer which stores a program permitting a computer to perform

a first process for specifying a modification relationship between a word to be predicted and a different word and/or word sequence, and

a second process for predicting said word using said different word and/or word sequence that is specified; and

a transmitter which reads said program from said storer and transmits said program.

19. A program storage device readable by machine, tangibly embodying a program of instructions executable by the machine to perform method steps for word prediction, said method comprising the steps of:

specifying a sentence structure consisting of multiple words, including a target word to be predicted;

employing said sentence structure to select a word and/or a word sequence that has a modification relationship with said target word to be predicted; and

predicting said target word based on said word and/or said word sequence that is selected.

20. A program storage device readable by machine, tangibly embodying a program of instructions executable by the machine to perform method steps for word prediction, said method comprising the steps of:

specifying a modification of a word to be predicted by a different word and/or word sequence; and

prediction said word using said different word and/or word sequence that is specified.

21. A program storage device readable by machine, tangibly embodying a program of instructions executable by the machine to perform method steps for speech recognition, said method comprising the steps of:

specifying the structure of a phrase from the beginning of a sentence to the j-th word (j=0, 1, 2, . . . );

employing a sentence structure up to said j-th word to specify one or multiple partial analysis trees modifying the (j+1)th word;

predicting said (j+1)th word based on said one or multiple partial analysis trees;

obtaining a sentence for a phrase including said predicted (j+1)th word and the probability value for said sentence structure; and

performing the above steps up to the last word of said sentence, and selecting, as speech recognition results, a sentence structure and a word sequence having maximum probability values.