Natural language speech lattice containing semantic variants

• US 7,873,519 B2
• Filed: 10/31/2007
• Issued: 01/18/2011
• Est. Priority Date: 11/12/1999
• Status: Expired due to Fees

• Current Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Original Assignee: Phoenix Solutions Incorporated (CDC Corporation)
• Inventors: Bennett, Ian
• Primary Examiner(s): Lerner, Martin
• Application Number:
  US11/932,684
• Publication Number:
  US 20080059153A1
• Time in Patent Office: 1,175 Days
• Field of Search: 704/9, 704/236, 704/238, 704/239, 704/255, 704/257, 704/270, 704/275, 704/243, 707/3, 707/4, 707/5, 707/708, 707/759, 707/771
• US Class Current: 704/257
• CPC Class Codes: G06F 16/24522 Translation of natural language queries to structured queries
• CPC Class Codes: G06F 16/3329 Natural language query formulation or dialogue systems
• CPC Class Codes: G06F 16/3344 using natural language analysis
• CPC Class Codes: G06F 40/216 using statistical methods
• CPC Class Codes: G06F 40/237 Lexical tools
• CPC Class Codes: G06F 40/289 Phrasal analysis, e.g. finite state techniques or chunking
• CPC Class Codes: G06F 40/30 Semantic analysis
• CPC Class Codes: G06F 40/42 Data-driven translation
• CPC Class Codes: G09B 19/06 Foreign languages with audible presentation of material to be studied G09B5/04
• CPC Class Codes: G09B 5/04 with audible presentation of the material to be studied sound recording or reproducing G11B
• CPC Class Codes: G09B 7/00 Electrically-operated teaching apparatus or devices working with questions and answers mechanically operated G09B3/00; computing arrangements G06F
• CPC Class Codes: G09B 7/02 of the type wherein the student is expected to construct an answer to the question which is presented or wherein the machine gives an answer to the question presented by a student
• CPC Class Codes: G10L 15/142 Hidden Markov Models [HMMs]
• CPC Class Codes: G10L 15/18 using natural language modelling
• CPC Class Codes: G10L 15/1815 Semantic context, e.g. disambiguation of the recognition hypotheses based on word meaning
• CPC Class Codes: G10L 15/183 using context dependencies, e.g. language models
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue
• CPC Class Codes: G10L 15/285 Memory allocation or algorithm optimisation to reduce hardware requirements
• CPC Class Codes: G10L 15/30 Distributed recognition, e.g. in client-server systems, for mobile phones or network applications
• CPC Class Codes: G10L 2015/223 Execution procedure of a spoken command
• CPC Class Codes: H04M 2250/74 : with voice recognition means voice activated dialling H04M1/271; voice recognition algorithms G10L15/00

Abstract

An intelligent query system for processing voiced-based queries is disclosed, which uses a combination of both statistical and semantic based processing to identify the question posed by the user by understanding the meaning of the user'"'"'s utterance. Based on identifying the meaning of the utterance, the system selects a single answer that best matches the user'"'"'s query. The answer that is paired to this single question is then retrieved and presented to the user. The system, as implemented, accepts environmental variables selected by the user and is scalable to provide answers to a variety and quantity of user-initiated queries.

Claims

1. A method of populating a natural language speech lattice with semantically variant questions, the method comprising the steps of:

(a) defining a first set of topic questions associated with content of a task domain to be supported by the natural language speech lattice;

wherein said first set of topic questions also each have at least one corresponding topic answer which can be provided by a speech based natural language system for the task domain in response to a speech based query;

(b) for each topic question;

i) dividing the topic question into a plurality of words corresponding to the topic question;

ii) determining semantically related words for each word in said plurality of words;

iii) formulating a semantic set of questions related to the topic question based on said semantically related words;

iv) performing semantic decoding on said semantic set of questions for the topic question, to identify a disambiguated set of questions appropriate for the topic question;

v) storing said set of disambiguated questions for the topic question in a speech recognition lattice;

(c) repeating step for each topic question in said first set of topic questions;

wherein said set of disambiguated questions for said first set of topic questions correspond to semantic variants of speech based queries that are supported by said speech based natural language system for the task domain.

2. The method of claim 1 wherein said semantically related words include at least one of a synonym, hyponym and/or hypernym.

3. The method of claim 1 wherein said semantically related words are determined from WORDNET.

4. The method of claim 1 wherein phrasal synonyms are also determined and used for said set of disambiguated questions.

5. The method of claim 1 wherein a statistical language model (SLM) is re-trained based on determining said set of disambiguated questions.

6. The method of claim 1 wherein said semantic decoding considers term frequency, coverage and semantic similarity to compute a semantic distance.

7. The method of claim 1 wherein said topic questions are automatically derived from transcriptions of user utterances.

8. The method of claim 1 wherein said topic questions are derived automatically by analyzing said content of said task domain.

9. The method of claim 1 further including a step:

embedding steps (a) through (c) as routines into a software data preparation tool.

10. The method of claim 1, further including a step:

processing a new user question and correlating the same to one or more of said set of disambiguated questions.

11. The method of claim 10 further including a step:

providing one or more answers to said new user question.

12. The method of claim 1 wherein said semantically related words are determined using a lexical dictionary.