System and method for speech enabled application

• US 6,094,635 A
• Filed: 09/17/1997
• Issued: 07/25/2000
• Est. Priority Date: 09/17/1997
• Status: Expired due to Term

• Current Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Original Assignee: Unisys Corporation
• Inventors: Scholz, Karl Wilmer, Blue, Reginald Victor, Diedrichs, Raymond Alan, Walsh, Joseph Patrick
• Primary Examiner(s): Hudspeth, David R.
• Assistant Examiner(s): Lerner, Martin
• Application Number:
  US08/932,938
• Time in Patent Office: 1,042 Days
• Field of Search: 704/8, 704/9, 704/10, 704/231, 704/243, 704/251, 704/257, 704/270, 704/275
• US Class Current: 704/270
• CPC Class Codes: G10L 15/1815 Semantic context, e.g. disambiguation of the recognition hypotheses based on word meaning
• CPC Class Codes: G10L 15/26 Speech to text systems G10L15/08 takes precedence

Abstract

The present invention is a computer apparatus and method for adding speech interpreting capabilities to an interactive voice response system. An annotated corpus is used to list valid utterances within a grammar along with token data for each valid utterance representing the meaning implied behind the valid utterance. When valid utterances are detected, the interactive voice response system requests that a search is made through the annotated corpus to find the token identified with the valid utterance. This token is returned to the interactive voice response system. If the valid utterance included a variable, additional processing is performed to interpret the variable and return additional data representing the variable.

Claims

1. A computer system for identifying the implied meaning behind a plurality of predetermined different valid spoken utterances in a grammar, said system comprising:

a. a central processing unit (CPU);

b. a system memory coupled to said CPU for receiving and storing memory files and applications files;

c. an automatic speech recognition (ASR) system coupled to said CPU comprising a speech recognizer coupled to receive spoken utterances and for producing valid utterances as an output as digital characters in word format;

d. a vendor specific automatic speech recognition (ASR) grammar stored in said system memory for cortrolling the output of said ASR system to provide said predetermined valid spoken utterances;

e. an interactive voice response (IVR) system coupled to said speech recognizer for receiving and processing said digital characters and words from said ASR;

f. runtime interpreter means stored in said system memory and coupled to said IVR system and said CPU for receiving from said IVR system said digital characters and words representative of said valid spoken utterances;

g. an annotated automatic speech recognition (ASR) corpus file, stored in said memory which contains a listing of all valid spoken utterances in said grammar and associated token data stored in said memory representing the implied meaning behind each said listed valid utterance; and

h. said runtime interpreter means comprising a runtime interpreter coupled between said ASR corpus file and said IVR system for receiving valid spoken utterances and searching said ASR corpus file for a match and for returning said token data representing said valid spoken utterance to said IVR system for a prompt or a response.

2. The system of claim 1, where said runtime interpreter means further includes means for performing a comparison search through said annotated ASR corpus in said to find said token data identifying said meaning of the implied system memory detected valid utterance.

3. The system of claim 2, where said runtime interpreter means further includes a routine interpretor for performing a partial match search through contents of said annotated ASR corpus upon failure of said comparison search, where said partial match search seeks a partial match of said detected valid utterance in said annotated ASR corpus.

4. The system of claim 3, where said runtime interpreter means further includes variable processing means for processing the unmatched portion of said detected valid utterance as a variable to identify the implied meaning of said unmatched portion.

5. The system of claim 4, where said variable processing means generates variable data representing the meaning of said unmatched portion of said detected valid utterance.

6. The system of claim 1, wherein said runtime interpreter means further comprising a runtime interpreter application program interface (RIAPI), coupled to said runtime interpreter, where said RIAPI is an interface used by said IVR system to access said runtime interpreter.

7. The system of claim 6, further comprising a custom processor (CP) interface, coupled to said RIAPI and said runtime interpreter, comprising an interface used by said RIAPI to access said runtime interpreter.

8. The computer system of claim 1, wherein said computer system is a network system having a plurality of said runtime interpreters distributed on a plurality of computers on said computer system network.

9. The computer system of claim 8, further comprising a resource manager for managing access to said plurality of runtime interpreters.

10. A speech enabled computer system, comprising:

a. a CPU;

b. a system memory coupled to said CPU;

c. a vendor-specific automatic speech recognition (ASR) grammar file stored in said system memory;

d. an automatic speech recognition system, coupled to said ASR grammar file in said CPU, for detecting predetermined valid spoken utterances and for generating as an output characters and words in digital format,

e. an annotated ASR corpus file, stored in said system memory, containing a listing of all predetermined valid utterances and token data representing the implied meaning of each listed valid utterance;

f. runtime interpreter means coupled to said annotated ASR corpus file and said CPU for searching through the contents of said annotated ASR corpus file to find token data representing the implied meaning of said detected valid spoken utterance;

g. said runtime interpreter means comprising a custom processor interface (CP), coupled to said CPU, for accessing aid corpus file;

h. a runtime interpreter application program interface (RIAPI), coupled to said CP; and

i. an interactive voice response system (IVR), coupled to said RIAPI and said ASR system, wherein said IVR issues requests to said RIAPI to search the contents of said annotated ASR corpus file and return token data representing the implied meaning of said valid spoken utterance detected by said ASR system.

11. A method for identifying the implied meaning behind a valid spoken utterance in a grammar, comprising the steps of:

a. loading an annotated ASR corpus file into a computer system memory, where said annotated ASR corpus file contains a listing of all various possible valid utterances in said grammar, as well as an associated token data for each of said listed valid utterances representing the implied meaning behind said valid utterances;

b. converting valid spoken utterances to digital requests;

c. receiving a request at an IVR to search said annotated ASR corpus file for the occurrence of a detected valid utterance;

d. performing a first search through said loaded annotated ASR corpus file in said system memory to find a said valid utterance using a runtime interpreter; and

e. returning token data corresponding to said detected valid utterance in said loaded annotated ASR corpus file to said sender of said request.

12. The method of claim 11, further comprising the steps of:

performing a second search through said loaded annotated ASR corpus upon failure of said first search, where said second search seeks a partial match of said detected valid utterance in said annotated ASR corpus; and

processing the unmatched portion of said detected valid utterance as a variable and returning variable data to the sender of said request, where said variable data represents the meaning of said unmatched portion.

13. The method of claim 11, further comprising the step of using a runtime interpreter application program interface (RIAPI) to access said ASR corpus file using said runtime interpreter.

14. The method of claim 13, further comprising the step of using a custom processor (CP) interface to access said runtime interpreter.

15. A computer system for identifying an implied meaning behind a plurality of different valid spoken utterances in a grammar and generating prompts and replies, said system comprising:

a) a central processing unit (CPU);

b) a system memory coupled to said CPU for receiving memory files and application files;

c) an automatic speech recognition (ASR) system coupled to receive said spoken utterances and for producing digital characters and words as an output;

d) a recognizer complier in said (ASR) system for establishing said plurality of valid spoken utterances as said ASR system output;

e) a vendor specific grammar in said CPU coupled to said ASR system;

f) an interactive voice response (IVR) system coupled to said automatic speech recognizer for receiving and processing said digital characters and words;

g) runtime interpreter means stored in said system memory of said CPU and coupled to said IVR system for receiving from said IVR system said digital characters and words representative of said valid spoken utterances;

h) an annotated automatic speech recognition (ASR) corpus file coupled to said runtime interpreter means and comprising a listing of all valid spoken utterances in a grammar format and associated token data representing the meaning behind each of said listed utterances; and

i) said runtime interpreter means being coupled between said ASR corpus file and said IVR systems for searching said ASR corpus file for a match and returning said associated token data to said IVR system to enable a prompt or a reply to said spoken utterances.

16. The system of claim 15 wherein said runtime interpreter means further includes an application program interface for coupling a known IVR system to said runtime interpreter.

17. The system of claim 16 wherein said ASR system comprises a known speech recognizer adapted for use in the present invention system by said recognizer complier for a vendor specific grammar.

18. The system as set forth in claim 15 wherein said annotated ASR corpus file further includes values of variables returned when a search in said ASR corpus file results in a partial match.

19. A natural language (NL) system for understanding the meaning behind sequences of predetermined different spoken words, said system having a central processor with a system memory for receiving and storing memory files and application files, said application files including an automatic speech recognition ASR system having a speech recognition (SR) grammar for recognizing valid spoken utterances, said NL system being characterized by;

an interactive voice response (IVR) system coupled to said ASR system for receiving predetermined valid spoken utterances recognized by said ASR system,

an ASR corpus file containing all combinations of valid spoken utterance recognizable by said ASR system and token data which represents the meaning behind each sequence of valid spoken utterances,

a runtime interpreter coupled between said IVR system and said ASR corpus file for searching said corpus file to find a matching valid utterance for each said spoken valid utterance, and for

obtaining said token data associated with said matching valid utterance, and for

returning said token data to said IVR to enable said IVR to generate a voice responsive prompt or a reply to said valid spoken utterance.