US5293584A
Expired / Lapsed
US
⏱️ 0.0 years remaining
Speech recognition system for natural language translation
Assignee: International Business Machines Corp
•
Filing: 1992-05-21
•
Granted: 1994-03-08
•
Expires: 2012-05-21
Abstract
Abstract A speech recognition system displays a source text of one or more words in a source language. The system has an acoustic processor for generating a sequence of coded representations of an utterance to be recognized. The utterance comprises a series of one or more words in a target language different from the source language. A set of one or more speech hypotheses, each comprising one or more words from the target language, are produced. Each speech hypothesis is modeled with an acoustic model. An acoustic match score for each speech hypothesis comprises an estimate of the closeness of a match between the acoustic model of the speech hypothesis and the sequence of coded representations of the utterance. A translation match score for each speech hypothesis comprises an estimate of the probability of occurrence of the speech hypothesis given the occurrence of the source text. A hypothesis score for each hypothesis comprises a combination of the acoustic match score and the translation match score. At least one word of one or more speech hypot This invention was made with Government support under Contract Number N00014-91-C-0135 awarded by the office of Naval Research. The Government has certain rights in this invention.
Inventors
- 👤 Peter F. Brown
- 👤 Stephen A. Della Pietra
- 👤 Vincent J. Della Pietra
- 👤 Frederick Jelinek
- 👤 Robert L. Mercer
Bibliographic Details
Application Number
—
Jurisdiction
US
Data Source
google_patents
Last Synchronized
2026-10-25 22:25
CPC Classifications
G; G10; G10L; G10L15/00; G10L15/08; G10L15/18
Claims Analysis
Total Claims: 38
Claim 1
Independent Claim
237 words
A speech recognition system comprising: means for displaying a source text comprising one or more words in a source language; an acoustic processor: for generating a sequence of coded representations of an utterance to be recognized, said utterance comprising one or more words in a target language different from the source language; means for generating a set of one or more speech hypotheses, each speech hypothesis comprising one or more words from the target language; means for generating an acoustic model of each speech hypothesis; means for generating an acoustic match score for each speech hypothesis, each acoustic match score comprising an estimate of the closeness of a match between the acoustic model of the speech hypothesis and the sequence of coded representations of the utterance; means for generating a translation match score for each speech hypothesis, each translation match score comprising an estimate of the probability of occurrence of the speech hypothesis given the occurrence, of the source text; means for generating a hypothesis score for each hypothesis, each hypothesis score comprising a combination of the acoustic match score and the translation match score for the hypothesis; means for storing a subset of one or more speech hypotheses, from the set of speech hypotheses, having the best hypothesis scores; and means for outputting at least one word of one or more of the speech hypotheses in the subset of speech hypotheses having the best hypothesis scores.
Claim 2
Dependent Claim
237 words
A speech recognition system comprising: means for displaying a source text comprising one or more words in a source language; an acoustic processor: for generating a sequence of coded representations of an utterance to be recognized, said utterance comprising one or more words in a target language different from the source language; means for generating a set of one or more speech hypotheses, each speech hypothesis comprising one or more words from the target language; means for generating an acoustic model of each speech hypothesis; means for generating an acoustic match score for each speech hypothesis, each acoustic match score comprising an estimate of the closeness of a match between the acoustic model of the speech hypothesis and the sequence of coded representations of the utterance; means for generating a translation match score for each speech hypothesis, each translation match score comprising an estimate of the probability of occurrence of the speech hypothesis given the occurrence, of the source text; means for generating a hypothesis score for each hypothesis, each hypothesis score comprising a combination of the acoustic match score and the translation match score for the hypothesis; means for storing a subset of one or more speech hypotheses, from the set of speech hypotheses, having the best hypothesis scores; and means for outputting at least one word of one or more of the speech hypotheses in the subset of speech hypotheses having the best hypothesis scores.
Claim 3
Dependent Claim
64 words
A speech recognition system as claimed in claim 1, characterized in that: the system further comprises means for generating a set of candidate words consisting solely of words in the target language which are partial or full translations of words in the source text; and the speech hypothesis generator generates one or more speech hypotheses solely from words in the set of candidate words.
Claim 4
Dependent Claim
52 words
A speech recognition system as claimed in claim 1, characterized in that the translation match score for a speech hypothesis comprises an estimate of the probability of occurrence of the source text given the occurrence of the speech hypothesis combined with an estimate of the probability of occurrence of the speech hypothesis.
Claim 5
Dependent Claim
57 words
A speech recognition system as claimed in claim 3, characterized in that the probability of occurrence of the source text given the occurrence of the speech hypothesis comprises an estimate, for each word in the source text, of the probability of the word in the source text given the occurrence of each word in the speech hypothesis.
Claim 6
Dependent Claim
40 words
A speech recognition system as claimed in claim 1, characterized in that the acoustic match score comprises an estimate of the probability of occurrence of the sequence of coded representations of the utterance given the occurrence of the speech hypothesis.
Claim 7
Dependent Claim
29 words
A speech recognition system as claimed in claim 5, characterized in that the hypothesis score comprises the product of the acoustic match score multiplied by the translation match score.
Claim 8
Dependent Claim
77 words
A speech recognition system as claimed in claim 1, further comprising: means for storing a source vocabulary of words in the source language; means for comparing each word in the source text with each word in the source vocabulary to identify each word in the source text which is not in the source vocabulary; and means for generating an acoustic model of at least one word in the source text which is not in the source vocabulary.
Claim 9
Dependent Claim
89 words
A speech recognition system as claimed in claim 7, characterized in that: each word in the source text has a spelling comprising one or more letters, each letter being upper case or being lower case; the system further comprises means for identifying each word in the source text which has an upper case first letter; and the means for generating an acoustic model generates an acoustic model of each word in the source text which is not in the source vocabulary, and which has an upper case first letter.
Claim 10
Dependent Claim
58 words
A speech recognition system as claimed in claim 7, characterized in that the means for generating an acoustic model comprises: means for storing a plurality of acoustic letter models; and means for generating an acoustic model of a word by replacing each letter in the spelling of the word with an acoustic letter model corresponding to the letter.
Claim 11
Dependent Claim
18 words
A speech recognition system as claimed in claim 1, characterized in that the output means comprises a display.
Claim 12
Dependent Claim
19 words
A speech recognition system as claimed in claim 10, characterized in that the display comprises a cathode ray tube.
Claim 13
Dependent Claim
19 words
A speech recognition system as claimed in claim 10, characterized in that the display comprises a liquid crystal display.
Claim 14
Dependent Claim
18 words
A speech recognition system as claimed in claim 1 characterized in that the output means comprises a printer.
Claim 15
Dependent Claim
18 words
A speech recognition system as claimed in claim 1, characterized in that the output means comprises a loudspeaker.
Claim 16
Dependent Claim
19 words
A speech recognition system as claimed in claim 1, characterized in that the output means comprises a speech synthesizer.
Claim 17
Dependent Claim
22 words
A speech recognition system as claimed in claim 1, characterized in that the means for storing speech hypotheses comprises readable computer memory.
Claim 18
Dependent Claim
196 words
A speech recognition system as claimed in claim 1, characterized in that the acoustic processor comprises: means for measuring the value of at least one feature of an utterance over each of a series of successive time intervals to produce a series of feature vector signals representing the feature values; means for storing a plurality of prototype vector signals, each prototype vector signal having at least one parameter value and having a unique identification value; means for comparing the closeness of the feature value of a first feature vector signal to the parameter values of the prototype vector signals to obtain prototype match scores for the first feature vector signal and each prototype vector signal; ranking means for associating a first-rank score with the prototype vector signal having the best prototype match score, and for associating a second-rank score with the prototype vector signal having the second best prototype match score; and means for outputting at least the identification value and the rank score of the first-ranked prototype vector signal, and the identification value and the rank score of the second-ranked prototype vector signal, as a coded utterance representation signal of the first feature vector signal.
Claim 19
Dependent Claim
29 words
A speech recognition system as claimed in claim 17, characterized in that the means for measuring the value of at least one feature of an utterance comprises a microphone.
Claim 20
Dependent Claim
217 words
A speech recognition method comprising: displaying a source text comprising one or more words in a source language; generating a sequence of coded representations of an utterance to be recognized, said utterance comprising one or more words in a target language different from the source language; generating a set of one or more speech hypotheses, each speech hypothesis comprising one or more words from the target language; generating an acoustic model of each speech hypothesis; generating an acoustic match score for each speech hypothesis, each acoustic match score comprising an estimate of the closeness of a match between the acoustic model of the speech hypothesis and the sequence of coded representations of the utterance; generating a translation match score for each speech hypothesis, each translation match score comprising an estimate of the probability of occurrence of the speech hypothesis given the occurrence of the source text; generating a hypothesis score for each hypothesis, each hypothesis score comprising a combination of the acoustic match score and the translation match score for the hypothesis; storing a subset of one or more speech hypotheses, from the set of speech hypotheses, having the best hypothesis scores; and outputting at least one word of one or more of the speech hypotheses in the subset of speech hypotheses having the best hypothesis scores.
Claim 21
Dependent Claim
217 words
A speech recognition method comprising: displaying a source text comprising one or more words in a source language; generating a sequence of coded representations of an utterance to be recognized, said utterance comprising one or more words in a target language different from the source language; generating a set of one or more speech hypotheses, each speech hypothesis comprising one or more words from the target language; generating an acoustic model of each speech hypothesis; generating an acoustic match score for each speech hypothesis, each acoustic match score comprising an estimate of the closeness of a match between the acoustic model of the speech hypothesis and the sequence of coded representations of the utterance; generating a translation match score for each speech hypothesis, each translation match score comprising an estimate of the probability of occurrence of the speech hypothesis given the occurrence of the source text; generating a hypothesis score for each hypothesis, each hypothesis score comprising a combination of the acoustic match score and the translation match score for the hypothesis; storing a subset of one or more speech hypotheses, from the set of speech hypotheses, having the best hypothesis scores; and outputting at least one word of one or more of the speech hypotheses in the subset of speech hypotheses having the best hypothesis scores.
Claim 22
Dependent Claim
67 words
A speech recognition method as claimed in claim 19, characterized in that: the method further comprises the step of generating a set of candidate words consisting solely of words in the target language which are partial or full translations of words in the source text; and the step of generating speech hypotheses generates one or more speech hypotheses solely from words in the set of candidate words.
Claim 23
Dependent Claim
52 words
A speech recognition method as claimed in claim 19, characterized in that the translation match score for a speech hypothesis comprises an estimate of the probability of occurrence of the source text given the occurrence of the speech hypothesis combined with an estimate of the probability of occurrence of the speech hypothesis.
Claim 24
Dependent Claim
57 words
A speech recognition method as claimed in claim 21, characterized in that the probability of occurrence of the source text given the occurrence of the speech hypothesis comprises an estimate, for each word in the source text, of the probability of the word in the source text given the occurrence of each word in the speech hypothesis.
Claim 25
Dependent Claim
40 words
A speech recognition method as claimed in claim 19, characterized in that the acoustic match score comprises an estimate of the probability of occurrence of the sequence of coded representations of the utterance given the occurrence of the speech hypothesis.
Claim 26
Dependent Claim
29 words
A speech recognition method as claimed in claim 23, characterized in that the hypothesis score comprises the product of the acoustic match score multiplied by the translation match score.
Claim 27
Dependent Claim
74 words
A speech recognition method as claimed in claim 19, further comprising the steps of: storing a source vocabulary of words in the source language; comparing each word in the source text with each word in the source vocabulary to identify each word in the source text which is not in the source vocabulary; and generating an acoustic model of at least one word in the source text which is not in the source vocabulary.
Claim 28
Dependent Claim
91 words
A speech recognition method is claimed in claim 25, characterized in that: each word in the source text bas a spelling comprising one or more letters, each letter being upper case or being lower case; the method further comprises the step of identifying each word in the source text which has an upper case first letter; and the step of generating an acoustic model comprises generating an acoustic model of each word in the source text which is not in the source vocabulary, and which has an upper case first letter.
Claim 29
Dependent Claim
54 words
A speech recognition method as claimed in claim 25, characterized in that the step of generating an acoustic model comprises: storing a plurality of acoustic letter models; and generating an acoustic model of a word by replacing each letter in the spelling of the word with an acoustic letter model corresponding to the letter.
Claim 30
Dependent Claim
18 words
A speech recognition method as claimed in claim 19, characterized in that the output means comprises a display.
Claim 31
Dependent Claim
19 words
A speech recognition method as claimed in claim 28, characterized in that the display comprises a cathode ray tube.
Claim 32
Dependent Claim
19 words
A speech recognition method as claimed in claim 28, characterized in that the display comprises a liquid crystal display,
Claim 33
Dependent Claim
18 words
A speech recognition method as claimed in claim 19, characterized in that the output means comprises a printer.
Claim 34
Dependent Claim
18 words
A speech recognition method as claimed in claim 19, characterized in that the output means comprises a loudspeaker.
Claim 35
Dependent Claim
19 words
A speech recognition method as claimed in claim 19, characterized in that the output means comprises a speech synthesizer.
Claim 36
Dependent Claim
22 words
A speech recognition method as claimed in claim 19, characterized in that the means for storing speech hypotheses comprises readable computer memory.
Claim 37
Dependent Claim
196 words
A speech recognition method as claimed in claim 19, characterized in that the acoustic processor comprises: means for measuring the value of at least one feature of an utterance over each of a series of successive time intervals to produce a series of feature vector signals representing the feature values; means for storing a plurality of prototype vector signals, each prototype vector signal having at least one parameter value and having a unique identification value; means for comparing the closeness of the feature value of a first feature vector signal to the parameter values of the prototype vector signals to obtain prototype match scores for the first feature vector signal and each prototype vector signal; ranking means for associating a first-rank score with the prototype vector signal having the best prototype match score, and for associating a second-rank score with the prototype vector signal having the second best prototype match score; and means for outputting at least the identification value and the rank score of the first-ranked prototype vector signal, and the identification value and the rank score of the second-ranked prototype vector signal, as a coded utterance representation signal of the first feature vector signal.
Claim 38
Dependent Claim
29 words
A speech recognition method as claimed in claim 35, characterized in that the means for measuring the value of at least one feature of in utterance comprises a microphone.
References Cited (Backward Citations)
US4443856A
Cited
US4809192A
Cited
US4613944A
Cited
US4507750A
Cited
US4987596A
Cited
US4903304A
Cited
US4980918A
Cited
US4748670A
Cited
US4759068A
Cited
US4852170A
Cited
US4896358A
Cited
US4882681A
Cited
US5072452A
Cited
US4984177A
Cited
US5146405A
Cited
EP0376501A2
Cited
US5062143A
Cited
USPTO Annuity & Maintenance Windows
| Annuity Stage | Window Opens | Payment Deadline | Surcharge Period Until | Estimated Fee | Status |
|---|---|---|---|---|---|
| 3.5 Years (Year 4) | 1997-03-07 | 1997-09-06 | 1998-03-08 | $2000.00 | Window Closed / Lapsed |
| 7.5 Years (Year 8) | 2001-03-07 | 2001-09-06 | 2002-03-08 | $3760.00 | Window Closed / Lapsed |
| 11.5 Years (Year 12) | 2005-03-07 | 2005-09-06 | 2006-03-08 | $7700.00 | Window Closed / Lapsed |