Method and subsystem for searching media content within a content-search service system

• US 9,015,172 B2
• Filed: 06/15/2012
• Issued: 04/21/2015
• Est. Priority Date: 09/22/2006
• Status: Expired due to Fees

• Current Assignee: Limelight Networks, Inc.
• Original Assignee: Limelight Networks, Inc.
• Inventors: Thompson, Jonathan, Chemburkar, Vijay, Bargeron, David, Acharya, Soam
• Primary Examiner(s): Chbouki, Tarek
• Application Number:
  US13/524,178
• Publication Number:
  US 20130013305A1
• Time in Patent Office: 1,040 Days
• US Class Current: 707/749
• CPC Class Codes: G06F 16/3335 Syntactic pre-processing, e.g. stopword elimination, stemming
• CPC Class Codes: G06F 16/367 Ontology
• CPC Class Codes: G06F 16/745 the internal structure of a single video sequence
• CPC Class Codes: G06F 16/78 Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
• CPC Class Codes: G06F 16/7844 using original textual content or text extracted from visual content or transcript of audio data

Abstract

Various embodiments of the present invention include concept-service components of content-search-service systems which employ ontologies and vocabularies prepared for particular categories of content at particular times in order to score transcripts prepared from content items to enable a search-service component of a content-search-service system to assign estimates of the relatedness of portions of a content item to search criteria in order to render search results to clients of the content-search-service system. The concept-service component processes a search request to generate lists of related terms, and then employs the lists of related terms to process transcripts in order to score transcripts based on information contained in the ontologies.

Claims

1. A concept-service component of a content-search-service system for searching a content item having an audio track, the concept-service component comprising:

a hardware processor configured to;

receive, as input, a content ID and search query, wherein the content ID uniquely identifies the content item;

use the content ID to retrieve a category ID, ontology, vocabulary, and a transcript, wherein;

the category ID relates to a subject matter of the content item, and

the transcript includes a textual rendering of the audio track;

correct and linguistically normalizes terms or phrases within the search query; and

use the linguistically normalized terms and phrases to process the transcript with a transcript scorer to assign a set of ontology-based scores, each ontology-based score of the set of ontology-based scores being associated with a different portion of the transcript, wherein the transcript-scorer;

prepares a list of term/ontology-metric pairs for each term or phrase in the linguistically normalized terms or phrases of the search query by;

identifying each entry in the ontology that includes the term or phrase paired with a second term; and

for each identified entry,

 

computing a co-occurrence metric as a combination of the co-occurrence values in the identified entry, and

 

adding an entry to the list that includes the second term and the computed co-occurrence metric; and

adding an entry to the list that includes the term and an identical-term co-occurrence metric; and

for each term or phrase in the transcript, associates a score with the term or phrase based on co-occurrence-metrics in the prepared lists of term/ontology-metric pairs; and

a memory coupled with the processor.

2. The concept-service component of the content-search-service system for searching the content item having the audio track as recited in claim 1 wherein the hardware processor is further configured to:

access a list of candidate tags for the content item;

establish a top concepts threshold value;

determine a score for each candidate tag based on a frequency or distribution of one or more words in the transcript, wherein the candidate tag comprises the one or more words or an ontology based on the candidate tag comprises the one or more words;

determine that the score for one or more candidate tags exceed the top concepts threshold value; and

associate information about the one or more candidate tags that exceed the top concepts threshold value with the content item to designate top concepts of the content item.

3. The concept-service component of the content-search-service system for searching the content item having the audio track as recited in claim 1 wherein the hardware processor is further configured to generate a second content item that selectively includes portions of the content item associated with above-threshold ontology based scores and not other portions of the content item.

4. The concept-service component of the content-search-service system for searching the content item having the audio track as recited in claim 1 wherein the search query is received from a user and is associated with a search across a set of content items, the set of content items comprising the content item.

5. The concept-service component of the content-search-service system for searching the content item having the audio track as recited in claim 1 wherein the transcript comprises a targeted transcript, the targeted transcript being sensitive to a plurality of words relative to other words, the plurality of words depending on the category ID.

6. The concept-service component of the content-search-service system for searching the content item having the audio track as recited in claim 1 wherein the resource-acquisition component requests a category ID and date or date/time indication that corresponds to the received content ID from a media-data-storage component of the content-search-service system.

7. The concept-service component of the content-search-service system for searching the content item having the audio track as recited in claim 6 wherein the resource-acquisition component requests an ontology and vocabulary from an ontology-store component of the content-search-service system, using the category ID and date or date/time indication and a transcript from an extracted-content-storage component using the received content ID.

8. The concept-service component of the content-search-service system for searching the content item having the audio track as recited in claim 1 wherein the search-query processor:

applies language rules and dictionary-based routines to the terms or phrases within the received search query to correct spellings of any misspelled terms in the search query;

applies language routines to normalize the terms or phrases within the received search query by changing plural forms to corresponding singular forms and replacing derivative terms with root forms of the derivative terms; and

filters from the search query terms that do not occur in the received vocabulary.

9. A method for searching for, and identifying, points in a transcribed media-content item related to a search query, the method comprising:

receiving, at a hardware processor, as input, a content ID and search query, wherein the content ID uniquely identifies a particular content item;

using the content ID to retrieve from a memory coupled with the hardware processor a category ID, ontology, vocabulary, and a transcript, wherein;

the category ID relates to a subject matter of the content item, and

the transcript includes a textual rendering of an audio track of the content item;

correcting and linguistically normalizing terms or phrases within the search query; and

using the linguistically normalized terms and phrases, at the hardware processor, to process the transcript to assign a set of ontology-based scores, each ontology-based score of the set of ontology-based scores being associated with a different portion of in the transcript, wherein using the linguistically normalized terms and phrases to process the transcript to assign a set of ontology-based scores comprises;

preparing a list of term/ontology-metric pairs for each term or phrase in the linguistically normalized terms or phrases of the search query; and

for each term or phrase in the transcript, associating a score with the term or phrase based on co-occurrence-metrics in the prepared lists of term/ontology-metric pairs by;

identifying each entry in each list of term/ontology-metric pairs in which the ontology that includes the currently considered term or phrase;

when two or more entries are identified, adding the co-occurrence metrics of the identified entries together and computing a score from the sum;

when one entry is identified, using the co-occurrence metric in the identified entry as the score; and

associating the score with the currently considered term or phrase.

10. The method for searching for, and identifying, points in the transcribed media-content item related to the search query as recited in claim 9, further comprising:

accessing a list of candidate tags for the content item;

establishing a top concepts threshold value;

determining a score for each candidate tag based on a frequency or distribution of one or more words in the transcript, wherein the candidate tag comprises the one or more words or an ontology based on the candidate tag comprises the one or more words;

determining that the score for one or more candidate tags exceed the top concepts threshold value; and

associating information about the one or more candidate tags that exceed the top concepts threshold value with the content item to designate top concepts of the content item.

11. The method for searching for, and identifying, points in the transcribed media-content item related to the search query as recited in claim 9 further comprising generating a second content item that selectively includes portions of the content item associated with above-threshold ontology based scores and not other portions of the content item.

12. The method for searching for, and identifying, points in the transcribed media-content item related to the search query as recited in claim 9 wherein the search query is received from a user and is associated with a search across a set of content items, the set of content items comprising the content item.

13. The method for searching for, and identifying, points in the transcribed media-content item related to the search query as recited in claim 9 wherein the transcript comprises a targeted transcript, the targeted transcript being sensitive to a plurality of words relative to other words, the plurality of words depending on the category ID.

14. The method for searching for, and identifying, points in the transcribed media-content item related to the search query as recited in claim 9 further comprising requesting a category ID and date or date/time indication that corresponds to the received content ID from a media-data-storage component of a content-search-service system.

15. The method for searching for, and identifying, points in the transcribed media-content item related to the search query as recited in claim 14 further comprising requesting an ontology and vocabulary from an ontology-store component of the content-search-service system, using the category ID and date or date/time indication and requesting a transcript from an extracted-content-storage component of the content-search-service system using the received content ID.

16. The method for searching for, and identifying, points in the transcribed media-content item related to the search query as recited in claim 9 wherein correcting and linguistically normalizing terms or phrases within the search query further comprises:

applying language rules and dictionary-based routines to the terms or phrases within the search query to correct spellings of any misspelled terms in the search query;

applying language routines to normalize the terms or phrases within the received search query by changing plural forms to corresponding singular forms and replacing derivative terms with root forms of the derivative terms; and

filtering from the search query terms that do not occur in the received vocabulary.

17. A machine-readable storage media having a set of instructions for searching for, and identifying, points in a transcribed media-content item related to a search query, the instructions, when executed by at least one machine, cause the at least one machine to:

receive, as input, a content ID and search query, wherein the content ID uniquely identifies a particular content item;

use the content ID to retrieve a category ID, ontology, vocabulary, and a transcript, wherein;

the category ID relates to a subject matter of the content item, and

the transcript includes a textual rendering of an audio track of the content item;

correct and linguistically normalize terms or phrases within the search query; and

use the linguistically normalized terms and phrases to process the transcript to assign a set of ontology-based scores, each ontology-based score of the set of ontology-based scores being associated with a different portion of in the transcript, wherein processing the transcript to assign a set of ontology-based scores comprises;

preparing a list of term/ontology-metric pairs for each term or phrase in the linguistically normalized terms or phrases of the search query, wherein preparing a list of term/ontology-metric pairs for each term or phrase in the linguistically normalized terms or phrases of the search query comprises;

identifying each entry in the ontology that includes the term or phrase paired with a second term; and

for each identified entry,

computing a co-occurrence metric as a combination of the co-occurrence values in the identified entry, and

adding an entry to the list that includes the second term and the computed co-occurrence metric; and

adding an entry to the list that includes the term and an identical-term co-occurrence metric; and

for each term or phrase in the transcript, associating a score with the term or phrase based on co-occurrence-metrics in the prepared lists of term/ontology-metric pairs.

18. The machine-readable storage media as recited in claim 17, wherein the instructions, when executed by the at least one machine, further cause the at least one machine to:

access a list of candidate tags for the content file;

establish a top concepts threshold value;

determine a score for each candidate tag based on a frequency or distribution of one or more words in the transcript, wherein the candidate tag comprises the one or more words or an ontology based on the candidate tag comprises the one or more words;

determine that the score for one or more candidate tags exceed the top concepts threshold value; and

associate information about the one or more candidate tags that exceed the top concepts threshold value with the content item to designate top concepts of the content item.

19. The machine-readable storage media as recited in claim 17, wherein the instructions, when executed by the at least one machine, further cause the at least one machine to generate a second content item that selectively includes portions of the content item associated with above-threshold ontology based scores and not other portions of the content item.

20. The machine-readable storage media as recited in claim 17 wherein the search query is received from a user and is associated with a search across a set of content items, the set of content items comprising the content item.

21. The machine-readable storage media as recited in claim 17 wherein the transcript comprises a targeted transcript, the targeted transcript being sensitive to a plurality of words relative to other words, the plurality of words depending on the category ID.

22. The machine-readable storage media as recited in claim 17 wherein the instructions, when executed by the at least one machine, further cause the at least one machine to request a category ID and date or date/time indication that corresponds to the received content ID from a media-data-storage component of a content-search-service system.

23. The machine-readable storage media as recited in claim 22 wherein the instructions, when executed by the at least one machine, further cause the at least one machine to request an ontology and vocabulary from an ontology-store component of the content-search-service system, use the category ID and date or date/time indication and request a transcript from an extracted-content-storage component of the content-search-service system using the received content ID.

24. The machine-readable storage media as recited in claim 17 wherein correcting and linguistically normalizing terms or phrases within the search query further comprises:

applying language rules and dictionary-based routines to the terms or phrases within the search query to correct spellings of any misspelled terms in the search query;

applying language routines to normalize the terms or phrases within the received search query by changing plural forms to corresponding singular forms and replacing derivative terms with root forms of the derivative terms; and

filtering from the search query terms that do not occur in the received vocabulary.