Method and Apparatus for Generation and Augmentation of Search Terms from External and Internal Sources

• US 20080104072A1
• Filed: 10/31/2007
• Published: 05/01/2008
• Est. Priority Date: 10/31/2002
• Status: Active Grant

• Current Assignee: Promptu Systems Corporation
• Original Assignee: Promptu Systems Corporation
• Inventors: Stampleman, Joseph, Printz, Harry
• Granted Patent: US 8,321,427 B2
• Time in Patent Office: Days
• US Class Current: 707/6
• CPC Class Codes: G06F 16/95 Retrieval from the web
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising
• CPC Class Codes: G10L 15/02 Feature extraction for speech recognition; Selection of recognition unit
• CPC Class Codes: G10L 15/142 Hidden Markov Models [HMMs]
• CPC Class Codes: G10L 15/18 using natural language modelling
• CPC Class Codes: G10L 15/187 Phonemic context, e.g. pronunciation rules, phonotactical constraints or phoneme n-grams
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue
• CPC Class Codes: G10L 17/26 Recognition of special voice characteristics, e.g. for use in lie detectors; Recognition of animal voices
• CPC Class Codes: G10L 2015/025 Phonemes, fenemes or fenones being the recognition units

Abstract

A method and apparatus to identify names, personalities, titles, and topics that are present in a repository, and place them into a grammar, and to identify names, personalities, titles, and topics that are not present in the repository, and place them into a grammar, uses information from external data sources, notably the text used in non-speech, text-based searches, to expand the search terms entered into the ASR grammars. The expansion takes place in two forms: (1) finding plausible linguistic variants of existing search terms that are already comprehended in the repository, but that are present under slightly different names; and (2) expanding the existing search term list with items that should be there by virtue of their currency in popular culture, but which for whatever reason have not yet been reflected with content items in the repository.

Claims

1. A method for identifying names, personalities, titles, and topics that may or may not be present in a repository and for placing them into a grammar, comprising the steps of:

using information from external data sources comprising non-speech, text-based searches to expand search terms entered into one or more automatic speech recognition ASR grammars;

said search terms expansion comprising any of;

finding plausible linguistic variants of existing search terms that are already comprehended in the repository, but that are under slightly different names; and

expanding an existing search term list with items that should be in said list by virtue of their currency in popular culture, but which for whatever reason have not yet been reflected with content items in the repository.

2. A method for identifying names, personalities, titles, and topics that may or may not be present in a repository and for placing them into a grammar for use in an automatic speech recognition (ASR) system, comprising the steps of:

extracting search term candidates from external sources;

extracting verified search terms from internal sources;

matching candidate search terms against verified search terms by edit distance techniques to obtain plausible linguistic variants of verified search terms;

using said linguistic variants to generate augmented verified search terms; and

including candidate search terms which do not point to actual content elements, but which the ASR system should nevertheless recognize, as null search terms in a final grammar by virtue of any of their high incidence count, repeated appearance in history as either a candidate or verified search term, or other criterion.

3. The method of claim 2, said external sources comprising any of:

published lists of frequent textual searches against popular search engines;

published lists of popular artists and songs; and

published lists of popular tags;

wherein candidate search terms are clearly identified as an explicitly marked title, author, artist name, etc. and, hence, processing is automatic; and

wherein human intervention is optionally used.

4. The method of claim 2, said external sources comprising any of:

published lists of most-emailed stories; and

published news feeds;

wherein a combination of automatic means, comprising either of named entity extraction (NEE) and/or topic detection and tracking (TDT) methods and direct human intervention are applied to running text or titles to generate candidate search terms.

5. The method of claim 2, said internal sources comprising any of:

explicitly marked titles, authors, artist names, etc. that are associated to content elements in said repository; and

sources obtained by application of named entity extraction (NEE) and/or topic detection and tracking (TDT) methods to descriptive text associated to content elements in said repository.

6. An apparatus for identifying names, personalities, titles, and topics that may or may not be present in a repository and for placing them into a grammar for use in an automatic speech recognition (ASR) system, comprising:

a repository comprising external sources and internal sources;

a module for combining data from all external sources and for generating an output comprising candidate search terms;

a module for processing said candidate search terms by performing any of incidence counting, low pass filtering, and other functions;

a module arranged to receive as inputs the output of said module for processing candidate search terms and historical information comprising any of a history of candidate search terms, a history of final search terms, and verified search terms, and for providing a processed external source output;

a module for receiving the processed external source output, and for identifying and outputting null search terms;

a module for combining data from all internal sources and for generating an output comprising said verified search terms;

an approximate text matching module arranged to receive as inputs said verified search terms and said candidate search terms, and for providing an output;

a module for receiving the output of said approximate text matching module and for providing an output comprising augmented verified search terms;

a module for combining said null search terms with said augmented verified search terms and for outputting final search terms for grammar generation.

7. The apparatus of claim 6, wherein said external sources comprise any of explicitly marked information and running text;

said external sources comprising;

an optional count filtering module for processing explicitly marked text, providing incidence count information is available, whereby only those instances with sufficiently high incidence count are retained; and

a module for processing running text by performing any of named entity extraction (NEE) and topic detection and tracking (TDT).

8. The apparatus of claim 6, wherein internal sources comprise any of explicitly marked information and running text;

said internal sources comprising a module for processing running text by performing any of named entity extraction (NEE) and topic detection and tracking (TDT).

9. An apparatus for identifying names, personalities, titles, and topics that may or may not be present in a repository and for placing them into a grammar, comprising:

a plurality of external data sources, comprising non-speech, text-based searches;

means for expanding search terms entered into one or more automatic speech recognition ASR grammars by using information from said external data sources, said means for expanding search terms comprising any of;

means for finding plausible linguistic variants of existing search terms that are already comprehended in the repository, but that are under slightly different names; and

means for expanding an existing search term list with items that should be in said list by virtue of their currency in popular culture, but which for whatever reason have not yet been reflected with content items in the repository.

10. The apparatus of claim 9, further comprising:

means for extracting search term candidates from said external sources;

means for extracting verified search terms from said internal sources;

means for matching candidate search terms against verified search terms by edit distance techniques to obtain plausible linguistic variants of verified search terms;

means for using said linguistic variants to generate augmented verified search terms; and

means for including candidate search terms which do not point to actual content elements, but which the ASR system should nevertheless recognize, as null search terms in a final grammar by virtue of any of their high incidence count, repeated appearance in history as either a candidate or verified search term, or other criterion.

11. The apparatus of claim 10, said external sources comprising any of:

published lists of frequent textual searches against popular search engines;

published lists of popular artists and songs; and

published lists of popular tags;

wherein candidate search terms are clearly identified as an explicitly marked title, author, artist name, etc. and, hence, processing is automatic; and

wherein human intervention is optionally used.

12. The apparatus of claim 10, said external sources comprising any of:

published lists of most-emailed stories; and

published news feeds;

wherein a combination of automatic means, comprising either of named entity extraction (NEE) and/or topic detection and tracking (TDT) methods and direct human intervention are applied to running text or titles to generate candidate search terms.

13. The apparatus of claim 10, said internal sources comprising any of:

explicitly marked titles, authors, artist names, etc. that are associated to content elements in said repository; and

sources obtained by application of named entity extraction (NEE) and/or topic detection and tracking (TDT) methods to descriptive text associated to content elements in said repository.

14. A method for identifying names, personalities, titles, and topics that may or may not be present in a repository and for placing them into a grammar for use in an automatic speech recognition (ASR) system, comprising the steps of:

providing a repository comprising said external sources and internal sources;

combining data from all external sources and generating candidate search terms;

processing said candidate search terms by performing any of incidence counting, low pass filtering, and other functions;

receiving said processed candidate search terms and historical information comprising any of a history of candidate search terms, a history of final search terms, and verified search terms, and providing a processed external source output;

receiving the processed external source output, and identifying and outputting null search terms;

combining data from all internal sources and generating said verified search terms;

receiving said verified search terms and said candidate search terms and providing an approximate text matching output;

receiving said approximate text matching output and outputting augmented verified search terms;

combining said null search terms with said augmented verified search terms and outputting final search terms for grammar generation.

15. The method of claim 14, wherein said external sources comprise any of explicitly marked information and running text;

said external sources subject to a process comprising the steps of;

processing explicitly marked text with an optional count filtering module providing incidence count information is available, whereby only those instances with sufficiently high incidence count are retained; and

processing running text by performing any of named entity extraction (NEE) and topic detection and tracking (TDT).

16. The method of claim 14, wherein internal sources comprise any of explicitly marked information and running text;

and

further comprising the step of processing running text by performing any of named entity extraction (NEE) and topic detection and tracking (TDT).