Generating context-based spell corrections of entity names

• US 10,162,895 B1
• Filed: 02/09/2015
• Issued: 12/25/2018
• Est. Priority Date: 03/25/2010
• Status: Active Grant

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: Google LLC (Alphabet Inc.)
• Inventors: Brunsman, Lawrence J., Devin, Matthieu, Lerner, Uri N., Tong, Simon
• Primary Examiner(s): Ly, Cheyne D
• Application Number:
  US14/616,915
• Time in Patent Office: 1,415 Days
• Field of Search: 707748, 707750, 707711, 707736
• CPC Class Codes: G06F 16/24 Querying
• CPC Class Codes: G06F 16/3338 Query expansion
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques
• CPC Class Codes: G06F 40/232 Orthographic correction, e.g. spell checking or vowelisation
• CPC Class Codes: G06F 40/247 Thesauruses; Synonyms
• CPC Class Codes: G06F 40/295 Named entity recognition

Abstract

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for correcting entity names. One method includes receiving texts and deriving a plurality of name-context pairs from the texts. The method further includes calculating a context consistency measure for each name-context pair and storing context-entity name data representing the name-context pairs. Another method includes identifying an entity name and one or more context terms from a query and generating candidate names for the entity name. The method further includes determining a score for each of the candidate names, selecting a number of top scoring candidate names, and using the selected candidate names to respond to the query.

Claims

1. A system comprising:

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

receiving a particular sequence of terms that occurs in a document, wherein the document includes multiple sequences of terms outside of the particular sequence;

determining that the particular sequence of terms includes one or more first terms that refer to a particular entity;

selecting one or more second terms from the particular sequence of terms that includes the one or more first terms that refer to the particular entity;

generating, for each of the selected one or more second terms that are from the particular sequence of terms that includes the one or more first terms that refer to the particular entity, a name-context pair that includes (i) the one or more first terms that refer to the particular entity, and (ii) the selected one or more second terms;

determining, for each distinct name-context pair, a context consistency measure that is an estimate of a probability that the selected one or more second terms of the respective name-context pair will appear in another sequence of terms based on the occurrence in the other sequence of terms of the one or more first terms that refer to the particular entity of the respective name-context pair; and

storing context-entity name data that associates one or more of the distinct name-context pairs with the corresponding context consistency measure in a memory.

2. The system of claim 1, wherein the context-entity name data maps each selected second term in the name-context pairs to a list of entity names and the context consistency measure for each entity name.

3. The system of claim 1, wherein the selected one or more second terms includes multiple consecutive words that occur before the one or more first terms that refer to the particular entity.

4. The system of claim 1, wherein the selected one or more second terms include multiple consecutive words that occur after the one or more first terms that refer to the particular entity.

5. The system of claim 1, wherein the operations comprise:

identifying a template for a document type of a document that includes the particular sequence of terms,

wherein the one or more second terms that are from the particular sequence of terms that includes the one or more first terms that refer to the particular entity are selected based at least on the template for the document type of the document that includes the particular sequence of terms.

6. The system of claim 1, wherein the operations comprise:

filtering the particular sequence of terms that includes the one or more first terms that refer to the particular entity to remove one or more of a stop word, a rare term, or a popular term.

7. The system of claim 1, wherein the operations comprise:

generating a score for each of the name-context pairs; and

excluding, from the context-entity name data, name-context pairs having a score that is less than a threshold.

19. The system of claim 1, wherein the particular sequence of terms that occurs in the document is a sentence.

20. The system of claim 1, wherein selecting one or more second terms from the particular sequence of terms that includes the one or more first terms that refer to the particular entity comprises:

selecting, as the one or more second terms, one or more terms that are adjacent to the one or more first terms that refer to the particular entity.

8. A computer-implemented method comprising:

receiving a particular sequence of terms that occurs in a document, wherein the document includes multiple sequences of terms outside of the particular sequence;

determining that the particular sequence of terms includes one or more first terms that refer to a particular entity;

selecting one or more second terms from the particular sequence of terms that includes the one or more first terms that refer to the particular entity;

generating, for each of the selected one or more second terms that are from the particular sequence of terms that includes the one or more first terms that refer to the particular entity, a name-context pair that includes (i) the one or more first terms that refer to the particular entity, and (ii) the selected one or more second terms;

determining, for each distinct name-context pair, a context consistency measure that is an estimate of a probability that the selected one or more second terms of the respective name-context pair will appear in another sequence of terms based on the occurrence in the other sequence of terms of the one or more first terms that refer to the particular entity of the respective name-context pair; and

storing context-entity name data that associates one or more of the distinct name-context pairs with the corresponding context consistency measure in a memory.

9. The method of claim 8, wherein the context-entity name data maps each selected second term in the name-context pairs to a list of entity names and the context consistency measure for each entity name.

10. The method of claim 8, wherein the selected one or more second terms includes multiple consecutive words that occur before the one or more first terms that refer to the particular entity.

11. The method of claim 8, wherein the selected one or more second terms include multiple consecutive words that occur after the one or more first terms that refer to the particular entity.

12. The method of claim 8, comprising:

identifying a template for a document type of a document that includes the particular sequence of terms,

wherein the one or more second terms that from the particular sequence of terms that includes the one or more first terms that refer to the particular entity are selected based at least on the template for the document type of the document that includes the particular sequence of terms.

13. The method of claim 8, comprising:

filtering the particular sequence of terms that includes the one or more first terms that refer to the particular entity to remove one or more of a stop word, a rare term, or a popular term.

14. The method of claim 8, comprising:

generating a score for each of the name-context pairs; and

excluding, from the context-entity name data, name-context pairs having a score that is less than a threshold.

15. A computer storage medium storing instructions that, when executed by data processing apparatus, cause the one or more computers to perform operations comprising:

receiving a particular sequence of terms that occurs in a document, wherein the document includes multiple sequences of terms outside of the particular sequence;

determining that the particular sequence of terms includes one or more first terms that refer to a particular entity;

selecting one or more second terms from the particular sequence of terms that includes the one or more first terms that refer to the particular entity;

generating, for each of the selected one or more second terms that are from the particular sequence of terms that includes the one or more first terms that refer to the particular entity, a name-context pair that includes (i) the one or more first terms that refer to the particular entity, and (ii) the selected one or more second terms;

determining, for each distinct name-context pair, a context consistency measure that is an estimate of a probability that the selected one or more second terms of the respective name-context pair will appear in another sequence of terms based on the occurrence in the other sequence of terms of the one or more first terms that refer to the particular entity of the respective name-context pair; and

storing context-entity name data that associates one or more of the distinct name-context pairs with the corresponding context consistency measure in a memory.

16. The medium of claim 15, wherein the context-entity name data maps each selected second term in the name-context pairs to a list of entity names and the context consistency measure for each entity name.

17. The medium of claim 15, wherein the selected one or more second terms includes multiple consecutive words that occur before the one or more first terms that refer to the particular entity.

18. The medium of claim 15, wherein the selected one or more second terms include multiple consecutive words that occur after the one or more first terms that refer to the particular entity.