Natural language processor for providing natural language signals in a natural language output

• US 10,394,963 B2
• Filed: 12/10/2015
• Issued: 08/27/2019
• Est. Priority Date: 10/22/2015
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Erickson, Thomas D., Farrell, Robert G.
• Primary Examiner(s): Jackson, Jakieda R
• Application Number:
  US14/965,368
• Publication Number:
  US 20170116185A1
• Time in Patent Office: 1,356 Days
• Field of Search: 704 9, 704 2
• CPC Class Codes: G06F 40/253 Grammatical analysis; Style critique
• CPC Class Codes: G06F 40/284 Lexical analysis, e.g. tokenisation or collocates
• CPC Class Codes: G06F 40/42 Data-driven translation
• CPC Class Codes: G06F 40/51 Translation evaluation
• CPC Class Codes: G06F 40/56 Natural language generation
• CPC Class Codes: G06F 40/58 Use of machine translation, e.g. for multi-lingual retrieval, for server-side translation for client devices or for real-time translation
• CPC Class Codes: G10L 15/1807 using prosody or stress

Abstract

Embodiments are directed to a natural language processing (NLP) system configured to receive a natural language (NL) input and perform an analysis operation to generate a NL output. The NLP system is configured to generate at least one confidence level based at least in part on at least one portion of the analysis operation. The NLP system is further configured to integrate at least one disfluency into the NL output based at least in part on the at least one confidence level.

Claims

1. A computer-implemented method of operating a speech synthesizer (SS) circuit configured to convert natural language inputs to natural language outputs and provide a natural language alert that communicates that the natural language outputs may contain error, the computer-implemented method comprising:

converting, using a machine translation circuit of the SS circuit, a natural language input to natural language input data, wherein the natural language input comprises a human source language;

performing, using the machine translation circuit of the SS circuit, a translation operation on the natural language input data to translate the natural language input data to a natural language output that represents a target human language;

wherein the translation operation comprises performing a confidence level analysis on at least one portion of the translation operation to generate at least one confidence level signal that represents a confidence level that the natural language output that results from the translation operation contains error;

wherein the translation operation further comprises, based at least in part on the at least one confidence level that the natural language output contains an error, selecting a portion of a disfluency natural language data stored in a memory and embedding the selected portion of the disfluency natural language data into the natural language output;

wherein the selected portion of the disfluency natural language data is embedded into the natural language output in a location selected to communicate that a portion of the natural language output may contain an error; and

converting, using the SS circuit, the natural language output that has been embedded with the selected portion of the disfluency natural language data into speech or text comprising a natural language output having disfluency, wherein the natural language output is in the target human language and the disfluency is in the target human language.

2. The computer-implemented method of claim 1 further comprising:

based at least in part on the confidence level, integrate at least one prosodic parameter into the natural language output data.

3. The computer-implemented method of claim 1, wherein the disfluency comprises a non-lexical utterance, which is derived from the target human language and communicates a degree of confidence.

4. The computer-implemented method of claim 1, wherein the speech comprises a linguistic utterance in the target human language that communicates a degree of confidence.

5. The computer-implemented method of claim 1, wherein:

the translation operation comprises applying first constraints on how the selected portion of the disfluency natural language data is embedded into the natural language output data; and

the translation operation further comprises applying second constraints on how the selected portion of the disfluency natural language data is integrated into the natural language output data.

6. The computer-implemented method of claim 5, wherein:

the first constraints comprise varying how the translation operation is performed; and

the second constraints comprise varying how the translation operation is performed.

7. A computer-implemented method of operating a speech synthesis (SS) system configured to convert natural language inputs to natural language outputs and provide a natural language alert that communicates that the natural language outputs may contain error, the computer-implemented method comprising:

receiving, at an input/output module of the SS processor, a natural language input comprising a natural language question;

converting, using the input/output module, the natural language input to natural language input data;

performing, using a processor of the SS system, an analysis operation on data of a corpus to generate natural language output data that represents a natural language answer to the natural language question;

wherein the analysis operation comprises performing a confidence level analysis on at least one portion of the analysis operation to generate at least one confidence level signal that represents a confidence level that the natural language output data that results from the analysis operation contains error;

wherein the analysis operation further comprises, based at least in part on the at least one confidence level that the natural language output contains an error, selecting a portion of a disfluency natural language data stored in a memory and embedding the selected portion of the disfluency natural language data into the natural language output;

wherein the selected portion of the disfluency natural language data is embedded into the natural language output in a location selected to communicate that a portion of the natural language output may contain an error; and

converting, using a speech synthesizer (SS) circuit of the natural language processor, the natural language output data that has been embedded with the selected portion of the disfluency natural language data into speech or text that comprises a natural language answer with a natural language disfluency.

8. The computer-implemented method of claim 7, wherein:

the disfluency comprises a non-lexical or a linguistic utterance that communicates a degree of confidence.

9. The computer-implemented method of claim 7, wherein:

the analysis operation comprises applying first constraints on how the selected portion of the disfluency natural language data is embedded into the natural language output data; and

the analysis operation comprises applying second constraints on how the selected portion of the disfluency natural language data is embedded into the natural language output data.

10. The computer-implemented method of claim 9, wherein:

the first constraints comprise varying how the analysis operation is performed; and

the second constraints comprise varying how the analysis operation is performed.