ENTITY NAME RECOGNITION

• US 20150371632A1
• Filed: 03/05/2015
• Published: 12/24/2015
• Est. Priority Date: 06/18/2014
• Status: Active Grant

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: Google Inc. (Alphabet Inc.)
• Inventors: Behzadi, Behshad, Skobeltsyn, Gleb
• Granted Patent: US 9,773,499 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G10L 15/1815 Semantic context, e.g. disambiguation of the recognition hypotheses based on word meaning
• CPC Class Codes: G10L 15/187 Phonemic context, e.g. pronunciation rules, phonotactical constraints or phoneme n-grams
• CPC Class Codes: G10L 15/26 Speech to text systems G10L15/08 takes precedence
• CPC Class Codes: G10L 2015/088 Word spotting
• CPC Class Codes: G10L 2015/228 of application context

Abstract

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for recognizing names of entities in speech. In one aspect, a method includes actions of receiving an utterance that includes (i) a first term that indicates a particular entity type, and (ii) a second term that indicates an entity name. Additional actions include obtaining a phonetic representation of the second term and determining that the phonetic representation of the second term matches a particular phonetic representation of a particular canonical name of a set of canonical names associated with a particular entity. Further actions include outputting a reference name associated with the particular entity as a transcription of the second term.

Claims

1. A computer-implemented method comprising:

accessing a directory of entities that indicates, for each entity, (i) a reference name of the entity, (ii) an entity type associated with the entity, and (iii) a geographic location associated with the entity;

obtaining, for each of one or more of the entities, a set of one or more canonical names for the entity, wherein the set of canonical names for the entity includes the reference name for the entity;

obtaining, for each of the one or more entities, a phonetic representation of each canonical name of the set of canonical names for the entity;

selecting a particular geographic area and a particular entity type;

selecting entities from the directory that (i) have an entity type that matches the particular entity type, and (ii) have a geographic location that matches the particular geographic area;

generating, for each of the selected entities, a record in an entity type-specific, geo-localized entity database, wherein, for each of the selected entities, the record indicates at least (i) a reference name of the selected entity, (ii) a respective phonetic representation of each canonical name of the set of canonical names for the selected entity;

receiving an utterance that includes (i) a first term that indicates the particular entity type, and (ii) a second term that indicates an entity name;

obtaining a phonetic representation of the second term;

determining that the phonetic representation of the second term matches a particular phonetic representation of a particular canonical name of a set of canonical names associated with a particular entity; and

in response to determining that the phonetic representation of the second term matches the particular phonetic representation of the particular canonical name of the set of canonical names associated with the particular entity, outputting the reference name associated with the particular entity as a transcription of the second term.

2. The method of claim 1, wherein obtaining, for each of the one or more entities, a phonetic representation of each canonical name of the set of canonical names for the entity, comprises:

generating the phonetic representations for each canonical name by applying one or more pronunciation rules to each canonical name.

3. The method of claim 1, wherein determining that the phonetic representation of the second term matches a particular phonetic representation of a particular canonical name of a set of canonical names associated with a particular entity, comprises:

determining a match score based on a distance between the phonetic representation of the second term and the particular phonetic representation; and

determining that the match score satisfies a predetermined threshold match score.

4. The method of claim 1, comprising:

determining a candidate transcription of the utterance, the candidate transcription including a type term corresponding to the first term and a name term corresponding to the second term;

determining a geographic location corresponding to the utterance; and

determining that the name term does not match a reference name of an entity (i) of the entity type corresponding to the type term and (ii) in a geographic location matching the geographic location,

wherein determining that the phonetic representation of the second term matches a particular phonetic representation of a particular canonical name of a set of canonical names associated with a particular entity is in response to determining that the name term does not match the reference name of the entity (i) of the entity type corresponding to the type term and (ii) in the geographic location matching the geographic location.

5. The method of claim 4, wherein determining a geographic location corresponding to the utterance, comprises:

determining a term in the candidate transcription that corresponds to the geographic location.

6. The method of claim 4, wherein determining a geographic location corresponding to the utterance, comprises:

determining a location in which utterance was said.

7. The method of claim 4, wherein obtaining a phonetic representation of the second term, comprises:

generating the phonetic representation of the second term by applying one or more pronunciation rules to the name term.

8. The method of claim 1, wherein a reference name comprises an official name and a canonical name comprises an official name or an unofficial name.

9. The method of claim 1, wherein an entity comprises at least one of a person, place, object, event, or concept.

10. A system comprising:

one or more computers; and

one or more storage devices storing instructions that are operable, when executed by the one or more computers, to cause the one or more computers to perform operations comprising;

accessing a directory of entities that indicates, for each entity, (i) a reference name of the entity, (ii) an entity type associated with the entity, and (iii) a geographic location associated with the entity;

obtaining, for each of one or more of the entities, a set of one or more canonical names for the entity, wherein the set of canonical names for the entity includes the reference name for the entity;

obtaining, for each of the one or more entities, a phonetic representation of each canonical name of the set of canonical names for the entity;

selecting a particular geographic area and a particular entity type;

selecting entities from the directory that (i) have an entity type that matches the particular entity type, and (ii) have a geographic location that matches the particular geographic area; and

generating, for each of the selected entities, a record in an entity type-specific, geo-localized entity database, wherein, for each of the selected entities, the record indicates at least (i) a reference name of the selected entity, (ii) a respective phonetic representation of each canonical name of the set of canonical names for the selected entity.

11. The system of claim 10, wherein obtaining, for each of the one or more entities, a phonetic representation of each canonical name of the set of canonical names for the entity, comprises:

generating the phonetic representations for each canonical name by applying one or more pronunciation rules to each canonical name.

12. The system of claim 10, wherein a reference name comprises an official name and a canonical name comprises an official name or an unofficial name.

13. The system of claim 10, wherein an entity comprises at least one of a person, place, object, event, or concept.

14. A computer-readable medium storing instructions executable by one or more computers which, upon such execution, cause the one or more computers to perform operations comprising:

receiving an utterance that includes (i) a first term that indicates a particular entity type, and (ii) a second term that indicates an entity name;

obtaining a phonetic representation of the second term;

determining that the phonetic representation of the second term matches a particular phonetic representation of a particular canonical name of a set of canonical names associated with a particular entity; and

in response to determining that the phonetic representation of the second term matches the particular phonetic representation of the particular canonical name of the set of canonical names associated with the particular entity, outputting a reference name associated with the particular entity as a transcription of the second term.

15. The medium of claim 14, wherein determining that the phonetic representation of the second term matches a particular phonetic representation of a particular canonical name of a set of canonical names associated with a particular entity, comprises:

determining a match score based on a distance between the phonetic representation of the second term and the particular phonetic representation; and

determining that the match score satisfies a predetermined threshold match score.

16. The medium of claim 14, the operations comprising:

determining a candidate transcription of the utterance, the candidate transcription including a type term corresponding to the first term and a name term corresponding to the second term;

determining a geographic location corresponding to the utterance; and

determining that the name term does not match a reference name of an entity (i) of the entity type corresponding to the type term and (ii) in a geographic location matching the geographic location,

wherein determining that the phonetic representation of the second term matches a particular phonetic representation of a particular canonical name of a set of canonical names associated with a particular entity is in response to determining that the name term does not match the reference name of the entity (i) of the entity type corresponding to the type term and (ii) in the geographic location matching the geographic location.

17. The medium of claim 16, wherein determining a geographic location corresponding to the utterance, comprises:

determining a term in the candidate transcription that corresponds to the geographic location.

18. The medium of claim 16, wherein determining a geographic location corresponding to the utterance, comprises:

determining a location in which utterance was said.

19. The medium of claim 16, wherein obtaining a phonetic representation of the second term, comprises:

generating the phonetic representation of the second term by applying one or more pronunciation rules to the name term.

20. The medium of claim 14, wherein a reference name comprises an official name and a canonical name comprises an official name or an unofficial name.