Disambiguation and tagging of entities

• US 8,533,223 B2
• Filed: 05/12/2009
• Issued: 09/10/2013
• Est. Priority Date: 05/12/2009
• Status: Active Grant

• Current Assignee: Comcast Interactive Media LLC (Comcast Corporation)
• Original Assignee: Comcast Interactive Media LLC (Comcast Corporation)
• Inventors: Houghton, David F.
• Primary Examiner(s): Breene, John E
• Assistant Examiner(s): ASPINWALL, EVAN S
• Application Number:
  US12/464,392
• Publication Number:
  US 20100293195A1
• Time in Patent Office: 1,582 Days
• Field of Search: 707/705, 707/706, 707/708, 707/723, 707/726, 707/730, 707/735, 707/738, 707/739, 707/740, 707/741, 707/748, 707/749, 707/750, 707/912, 707/913, 707/914, 707/916, 707/943, 707/759, 707/765, 707/776
• US Class Current: 707/776
• CPC Class Codes: G06F 16/285 Clustering or classification
• CPC Class Codes: G06F 40/169 Annotation, e.g. comment data or footnotes
• CPC Class Codes: G06F 40/284 Lexical analysis, e.g. tokenisation or collocates
• CPC Class Codes: G06F 40/289 Phrasal analysis, e.g. finite state techniques or chunking
• CPC Class Codes: G06F 40/295 Named entity recognition
• CPC Class Codes: G06F 40/30 Semantic analysis

Abstract

Tagging of content items and entities identified therein may include a matching process, a classification process and a disambiguation process. Matching may include the identification of potential matching candidate entities in a content item whereas the classification process may categorize or group identified candidate entities according to known entities to which they are likely a match. In some instances, a candidate entity may be categorized with multiple known entities. Accordingly, a disambiguation process may be used to reduce the potential matches to a single known entity. In one example, the disambiguation process may include ranking potentially matching known entities according to a hierarchy of criteria.

Claims

1. A method comprising:

identifying, at a processor, a candidate entity in a media content item, wherein the candidate entity is a potential identification of a first known entity, wherein the media content item comprises a plurality of words or phrases and wherein the candidate entity corresponds to a first word or first phrase of the plurality of words or phrases;

performing a first categorization of the candidate entity based on a first set of factors, wherein the first categorization comprises categorizing at least one other candidate entity following the candidate entity in the media content item according to a sequence in which the media content item is to be consumed by a user, wherein the at least one other candidate entity corresponds to a second word or second phrase of the plurality of words or phrases;

performing a second categorization of the candidate entity, after the first categorization, based on a second set of factors, wherein the second set of factors is different from the first set of factors and comprises, at least in part, results of the first categorization associated with the at least one other candidate entity;

determining, after the second categorization, that the candidate entity is categorized with a plurality of known entities;

in response to determining that the candidate entity is categorized with the plurality of known entities, selecting one of the plurality of known entities corresponding to the candidate entity, and

tagging the candidate entity, based on the selection of the one of the plurality of known entities and separately from tagging the at least one other candidate entity, as identifying the selected one of the plurality of known entities.

2. The method of claim 1, wherein identifying the candidate entity in the media content item is performed using string matching.

3. The method of claim 1, wherein the candidate entity includes at least one of:

a person'"'"'s name and a title of a content item.

4. The method of claim 1, wherein the first set of factors includes a rate of co-occurrence between the candidate entity and the at least one other candidate entity.

5. The method of claim 1, wherein tagging the candidate entity includes assigning, in the memory, a database identifier of one of the plurality of known entities to the candidate entity.

6. The method of claim 1, wherein the first set of factors includes a word that immediately follows the candidate entity in the media content item.

7. The method of claim 1, wherein the plurality of known entities includes the first known entity and a second known entity and wherein selecting one of the plurality of known entities corresponding to the candidate entity includes:

ranking the first known entity and the second known entity based on a first categorization reliability criterion;

determining that the first known entity is ranked lower than the second known entity; and

removing the first known entity as a potential identification of the candidate entity.

8. The method of claim 1, further comprising associating the tagged candidate entity with a link to additional information about the candidate entity.

9. The method of claim 1, further comprising:

discarding the at least one other candidate entity as a potential identification of the first known entity during the second categorization;

determining, after completing the first and second categorizations, that a set of one or more candidate entities in the media content item are categorized as identifying the first known entity; and

in response to determining that the set of one or more candidate entities are categorized as identifying the first known entity, adding the at least one other candidate entity to the set of one or more candidate entities categorized as identifying the first known entity.

10. The method of claim 1, wherein the plurality of words or phrases are identified from at least one of:

text, an image and speech.

11. The method of claim 7, wherein the plurality of known entities includes a third known entity and wherein selecting one of the plurality of known entities corresponding to the candidate entity further includes:

ranking the third known entity and the second known entity based on a second categorization reliability criterion;

determining that the third known entity is ranked lower than the second known entity based on the second match categorization criterion; and

removing the third known entity as a potential identification of the candidate entity.

12. The method of claim 7, wherein the first categorization reliability criterion includes a number of candidate entities in the media content item categorized as potential identifications of the first known entity or the second known entity.

13. The method of claim 7, wherein the first categorization reliability criterion includes a categorization confidence.

14. The method of claim 7, wherein the first categorization reliability criterion includes a co-occurrence rate between the candidate entity and another candidate entity that is categorized with only one known entity.

15. One or more non-transitory computer readable media storing computer readable instructions that, when executed, cause an apparatus to:

identify, at the apparatus, a candidate entity in a media content item, wherein the candidate entity is a potential identification of a first known entity wherein the media content item comprises a plurality of words or phrases and wherein the candidate entity corresponds to a first word or first phrase of the plurality of words or phrases;

perform a first categorization of the candidate entity based on a first set of factors, wherein the first categorization comprises categorizing at least one other candidate entity following the candidate entity in the media content item according to a sequence in which the media content is configured to be consumed by a user, wherein the at least one other candidate entity corresponds to a second word or second phrase of the plurality of words or phrases;

perform a second categorization of the candidate entity after the first categorization based on a second set of factors, wherein the second set of factors is different from the first set of factors and comprises at least in part, results of the first categorization associated with the at least one other candidate entity;

determine, after the second categorization, that the candidate entity is categorized with a plurality of known entities;

in response to determining that the candidate entity is categorized with a plurality of known entities, select one of the plurality of known entities corresponding to the candidate entity; and

tag the candidate entity, based on the selection of the one of the plurality of known entities and separately from tagging the at least one other candidate entity, as identifying the selected one of the plurality of known entities.

16. The one or more computer readable media of claim 15, wherein the first set of factors includes a word that immediately follows the candidate entity in the content item.

17. The one or more computer readable media of claim 15, wherein the plurality of known entities includes the first known entity and a second known entity and wherein selecting one of the plurality of known entities corresponding to the candidate entity includes:

ranking the first known entity and the second known entity based on a first categorization reliability criterion;

determining that the first known entity is ranked lower than the second known entity; and

removing the first known entity as a potential match with the candidate entity.

18. A method comprising:

receiving a sequence of words or phrases occurring in a piece of video, audio, audiovisual, or textual content;

sequentially processing the sequence of words or phrases in a first categorization process, wherein the first categorization process includes categorizing each of a first word or phrase and a second word or phrase in the sequence as identifying one or more corresponding known entities, wherein the second word or phrase follows the first word or phrase in the sequence, and wherein categorizing the first word or phrase during the first categorization process is performed without using a categorization of the second word or phrase;

processing the sequence of words or phrases in a second categorization process including performing another categorization of the first word or phrase using the categorization of the second word or phrase resulting from the first categorization process;

upon determining that the first word or phrase is categorized with a plurality of known entities, selecting one of the plurality of known entities as being identified by the first word or phrase; and

tagging the first word or phrase as identifying the selected one of the plurality of known entities.

19. An apparatus comprising:

a processor; and

memory storing computer readable instructions that, when executed, cause the apparatus to;

identify a candidate entity in a media content item, wherein the candidate entity is a potential identification of a first known entity, wherein the media content item comprises a plurality of words or phrases and wherein the candidate entity corresponds to a first word or first phrase of the plurality of words or phrases;

perform a first categorization of the candidate entity based on a first set of factors, wherein the first categorization comprises categorizing at least one other candidate entity following the candidate entity in the media content item according to a sequence in which the media content is configured to be consumed by a user, wherein the at least one other candidate entity corresponds to a second word or second phrase of the plurality of words or phrases;

perform a second categorization of the candidate entity, after the first categorization, based on a second set of factors, wherein the second set of factors is different from the first set of factors and comprises, at least in part, results of the first categorization associated with the at least one other candidate entity;

determine, after the second categorization, that the candidate entity is categorized with a plurality of known entities;

in response to determining that the candidate entity is categorized with the plurality of known entities, select one of the plurality of known entities corresponding to the candidate entity; and

tag the candidate entity, based on the selection of the one of the plurality of known entities and separately from tagging the at least one other candidate entity, as identifying the selected one of the plurality of known entities.

20. An apparatus comprising:

a processor; and

memory storing computer readable instructions that, when executed, cause the apparatus to;

receive a sequence of words or phrases occurring in a piece of video, audio, audiovisual, or textual content;

sequentially process the sequence of words or phrases in a first categorization process, wherein the first categorization process includes categorizing each of a first word or phrase and a second word or phrase in the sequence as identifying one or more corresponding known entities, wherein the second word or phrase follows the first word or phrase in the sequence, and wherein categorizing the first word or phrase during the first categorization process is performed without using a categorization of the second word or phrase;

process the sequence of words or phrases in a second categorization process including performing another categorization of the first word or phrase using the categorization of the second word or phrase resulting from the first categorization process;

upon determining the first word or phrase is categorized with a plurality of known entities, select one of the plurality of known entities as being identified by the first word or phrase; and

tag the first word or phrase as identifying the selected one of the plurality of known entities.

21. One or more non-transitory computer readable media storing computer readable instructions that, when executed, cause an apparatus to:

receive a sequence of words or phrases occurring in a piece of video, audio, audio visual or textual content;

sequentially process the sequence of words or phrases in a first categorization process, wherein the first categorization process includes categorizing each of a first word or phrase and a second word or phrase in the sequence as identifying one or more corresponding known entities, wherein the second word or phrase follows the first word or phrase in the sequence, and wherein categorizing the first word or phrase during the first categorization process is performed without using a categorization of the second word or phrase;

process the sequence of words or phrases in a second categorization process including performing another categorization of the first word or phrase using the categorization of the second word or phrase resulting from the first categorization process;

upon determining that the first word or phrase is categorized with a plurality of known entities, select one of the plurality of known entities as being identified by the first word or phrase; and

tag the first word or phrase as identifying the selected one of the plurality of known entities.