Method of automated discovery of topics relatedness

• US 9,542,477 B2
• Filed: 12/02/2014
• Issued: 01/10/2017
• Est. Priority Date: 12/02/2013
• Status: Active Grant

• Current Assignee: Finch Computing LLC (Qbase, LLC)
• Original Assignee: Qbase, LLC
• Inventors: Lightner, Scott, Weckesser, Franz, Boddhu, Sanjay, Dave, Rakesh, Flagg, Robert
• Primary Examiner(s): ALAM, SHAHID AL
• Application Number:
  US14/557,906
• Publication Number:
  US 20150154305A1
• Time in Patent Office: 770 Days
• Field of Search: 707/726, 707/791
• US Class Current: 1/1
• CPC Class Codes: G06F 16/10 File systems; File servers
• CPC Class Codes: G06F 16/3334 Selection or weighting of terms from queries, including natural language queries
• CPC Class Codes: G06F 16/3347 using vector based model
• CPC Class Codes: G06F 16/93 Document management systems
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation
• CPC Class Codes: G06F 16/9538 Presentation of query results

Abstract

A computer system and method for automated discovery of topic relatedness are disclosed. According to an embodiment, topics within documents from a corpus may be discovered by applying multiple topic identification (ID) models, such as multi-component latent Dirichlet allocation (MC-LDA) or similar methods. Each topic model may differ in a number of topics. Discovered topics may be linked to the associated document. Relatedness between discovered topics may be determined by analyzing co-occurring topic IDs from the different models, assigning topic relatedness scores, where related topics may be used for matching/linking a feature of interest. The disclosed method may have an increased disambiguation precision, and may allow the matching and linking of documents using the discovered relationships.

Claims

1. A method comprising:

generating, via a first topic model computer, a first term vector identifying a first topic in a plurality of documents in a document corpus;

generating, via a second topic model computer, a second term vector identifying a second topic in the plurality of documents in the document corpus;

linking, via a topic detection computer, each of the first and second topics across the plurality of documents in the document corpus, wherein linking comprises matching of the each of the first and second topics across the plurality of documents in the document corpus and indicates a tag associated with metadata that the first and second topics are each identified in at least one document in the document corpus;

assigning, via the topic detection computer, a relatedness score weight to each of the linked first and second topics based on co-occurrence of each of the linked first and second topics across the plurality of documents in the document corpus;

determining, via the topic detection computer, whether the first and second linked topics are related across the plurality of documents in the document corpus based at least in part on the relatedness score weight;

executing via the first topic model computer, a master topic computer model based on a multi-component extension of latent Dirichlet allocation having a first set of model parameters; and

executing via the second topic model computer, a periodic new topic computer model based on the multi-component extension of latent Dirichlet allocation having a second set of model parameters different from the first set of model parameters.

2. The method of claim 1, wherein the topic detection computer determines one or more differences between the first and second topics to identify at least one new topic and adds the at least one new topic as a model parameter to the first topic model computer.

3. The method of claim 1, further comprising generating, via the change detection computer, a hierarchy of related topics in the document corpus.

4. The method of claim 1, wherein the first topic model computer and the second topic model computer execute respective first and second topic computer models having one or more parameters selected from the group consisting of a multi-document component, a vocabulary size, and a parameter setting for a prior Dirichlet distribution on a topic term.

5. The method of claim 1, wherein the linking step further comprising generating a graphical representation of co-occurring linked topics across the plurality of documents in the document corpus.

6. A system comprising:

a first computer comprising a processor executing a master topic model (MTM) computer module, the first computer configured to generate a first term vector identifying a first topic in a plurality of documents in a document corpus;

a second computer comprising a processor executing a periodic new model (PNM) computer module, the second computer configured to generate a second term vector identifying a second topic in the plurality of documents in the document corpus; and

a third computer comprising a processor executing a change detection computer module, the third computer configured to;

(a) link each of the first and second topics across the plurality of documents in the document corpus by matching the first and second topics across the plurality of documents in the document corpus, a link indicating a tag associated with metadata that the first and second topics are each identified in at least one document in the document corpus,

(b) assign a relatedness score weight to each of the linked first and second topics based on co-occurrence of each of the linked first and second topics across the plurality of documents in the document corpus,

(c) determine whether the first and second linked topics are related across the plurality of documents in the document corpus based at least in part on the relatedness score weight;

(d) execute a master topic computer model based on a multi-component extension of latent Dirichlet allocation having a first set of model parameters; and

the second computer'"'"'s processor further executing a periodic new topic module to detect a new topic, the new topic module configuring the second computer to perform a new topic model based on the multi-component extension of latent Dirichlet allocation having a second set of model parameters different from the first set of model parameters.

7. The system of claim 6, wherein the third computer is further configured to determine one or more differences between the first and second topics to identify at least one new topic and add the at least one new topic as a model parameter to the first computer.

8. The system of claim 6, wherein the third computer is further configured to generate a hierarchy of related topics in the document corpus.

9. The system of claim 6, wherein the first computer and the second computer are further configured to execute respective first and second topic computer models having one or more parameters selected from the group consisting of a multi-document component, a vocabulary size, and a parameter setting for a prior Dirichlet distribution on a topic term.

10. The system of claim 6, wherein the change detection computer module is further configured to generate a graphical representation of co-occurring linked topics across the plurality of documents in the document corpus.

11. A non-transitory computer readable medium having stored thereon computer executable instructions comprising:

generating, via a first topic model computer module of a computer, a first term vector identifying a first topic in a plurality of documents in a document corpus;

generating, via a second topic model computer module of the computer, a second term vector identifying a second topic in the plurality of documents in the document corpus;

linking, via a topic detection computer module of the computer, each of the first and second topics across the plurality of documents in the document corpus, wherein linking comprises matching of the each of the first and second topics across the plurality of documents in the document corpus, and indicates a tag associated with metadata that the first and second topics are each identified in at least one document in the document corpus;

assigning, via the topic detection computer module of the computer, a relatedness score weight to each of the linked first and second topics based on co-occurrence of each of the linked first and second topics across the plurality of documents in the document corpus;

determining, via the topic detection computer module of the computer, whether the first and second linked topics are related across the plurality of documents in the document corpus based at least in part on the relatedness score weight

executing, via the first topic model computer module, a master topic computer model based on a multi-component extension of latent Dirichlet allocation having a first set of model parameters; and

executing, via the second topic model computer module, a periodic new topic computer model based on the multi-component extension of latent Dirichlet allocation having a second set of model parameters different from the first set of model parameters.

12. The computer readable medium of claim 11 wherein the instructions further comprise determining, via the topic detection computer module of the computer, one or more differences between the first and second topics to identify at least one new topic and adds the at least one new topic as a model parameter to the first topic model computer module.

13. The computer readable medium of claim 11 wherein the instructions further comprise generating, via the topic detection computer module of the computer, a hierarchy of related topics in the document corpus.

14. The computer readable medium of claim 11 wherein the instructions further comprise executing, via the first and second topic model computer modules, respective first and second topic computer models having one or more parameters selected from the group consisting of a multi-document component, a vocabulary size, and a parameter setting for a prior Dirichlet distribution on a topic term.