KNOWLEDGE DISCOVERY FROM CITATION NETWORKS

• US 20160171391A1
• Filed: 02/22/2016
• Published: 06/16/2016
• Est. Priority Date: 12/06/2010
• Status: Active Grant

• Current Assignee: The Research Foundation for The State University of New York (State University of New York)
• Original Assignee: The Research Foundation for The State University of New York (State University of New York)
• Inventors: Guo, Zhen, Zhang, Mark
• Granted Patent: US 9,892,367 B2
• Time in Patent Office: Days
• CPC Class Codes: G06F 16/24573 using data annotations, e.g. user-defined metadata
• CPC Class Codes: G06F 16/24578 using ranking
• CPC Class Codes: G06F 16/93 Document management systems
• CPC Class Codes: G06F 16/9558 Details of hyperlinks; Management of linked annotations
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 7/01 Probabilistic graphical models, e.g. probabilistic networks
• CPC Class Codes: G06Q 10/10 Office automation; Time management

Abstract

In a corpus of scientific articles such as a digital library, documents are connected by citations and one document plays two different roles in the corpus: document itself and a citation of other documents. A Bernoulli Process Topic (BPT) model is provided which models the corpus at two levels: document level and citation level. In the BPT model, each document has two different representations in the latent topic space associated with its roles. Moreover, the multi-level hierarchical structure of the citation network is captured by a generative process involving a Bernoulli process. The distribution parameters of the BPT model are estimated by a variational approximation approach.

Claims

1. A method for processing a corpus of documents having a multi-level transitive linkage structure with other documents, comprising:

providing a computer-implemented generative model based on at least parameters estimated by probabilistic inference, which models a content of each document in the document database based on at least

(i) an intrinsic content of each respective document; and

(ii) a content of related documents linked to each respective document through the multi-level transitive linkage structure,

the computer-implemented generative model representing the content of each respective document as at least a mixture over latent topics having topic distributions which are a mixture of distributions associated with the related documents comprising at least a mixture weighting of the intrinsic content of each respective document and the content of related documents linked to each respective document through the multi-level transitive linkage structure; and

at least one of representing, characterizing, clustering, summarizing, indexing, ranking, and searching the documents in the corpus of documents with at least one automated processor based on at least the computer-implemented generative model.

2. The method according to claim 1, further comprising receiving a user input, and determining, with at least one automated processor, at least one document from the corpus of documents selectively associated with the received user input.

3. The method according to claim 2, further comprising ranking a plurality of documents of the corpus of documents in dependence on the received user input and the computer-implemented generative model.

4. The method according to claim 1, wherein the multi-level transitive citation structure comprises at least one of citations, authors, cross references, and hyperlinks.

5. The method according to claim 1, wherein the generative model is a Bernoulli Process Topic model.

6. The method according to claim 1, wherein the multi-level structure of the citation network is captured by a Bernoulli process.

7. The method according to claim 1, wherein relations among the documents of the corpus of documents at a document level are represented as mixtures of the topic distributions at a document citation level.

8. The method according to claim 1, wherein the latent topics are distinguished by Bayesian inference.

9. The method according to claim 1, further comprising receiving a query as an input, and generating a search output responsive to the query selectively dependent on the computer-implemented generative model.

10. The method according to claim 1, further comprising generating a visualized output dependent on the at least one of representing, characterizing, clustering, summarizing, indexing, ranking, and searching of the documents in the corpus of documents.

11. The method according to claim 1, further comprising, for each location within each document of the corpus of documents:

choosing a related document from a multinomial distribution conditioned on each document,

choosing a topic from the topic distribution of the linked document at the related document level, and

choosing a word which follows a multinomial distribution conditioned on the chosen topic.

12. A method for processing a set of documents having a multi-level transitive linkage structure with other documents within the set of documents, comprising:

generating a probabilistic inference model of a content of each document of the set of documents, based on at least;

(i) an intrinsic content of each document,

(ii) a content of related documents linked to each document through the multi-level transitive linkage structure; and

(iii) a mixture weighting of the intrinsic content of each document and the content of the related documents linked to each document through the multi-level transitive linkage structure, to represent the content of each document as at least a mixture over latent topics having topic distributions which are a mixture of distributions associated with the related documents linked to each document through the multi-level transitive linkage structure; and

at least one of representing, characterizing, clustering, summarizing, indexing, ranking, and searching the documents in the corpus of documents with at least one automated processor based on at least the computer-implemented generative model.

13. A system for processing a corpus of documents having a multi-level transitive linkage structure with other documents, comprising:

a memory configured to store a computer-implemented generative model dependent on at least parameters estimated by probabilistic inference, which models a content of each respective document in the document database as at least a mixture weighting of an intrinsic content of each respective document and a content of documents related to the respective document through multi-level transitive linkage structure, to represent the content of each respective document as at least a mixture over latent topics having topic distributions which are a mixture of distributions associated with documents related to the respective document through multi-level transitive linkage structure; and

at least one automated processor configured to at least one of represent, characterize, cluster, summarize, index, rank, and search the documents in the corpus of documents with at least one automated processor based on at least the computer-implemented generative model.

14. The system according to claim 13, further comprising a user input configured to receive information, wherein the at least one automated processor is further configured to select at least one document from the corpus of documents selectively associated with the received user input.

15. The system according to claim 14, wherein the at least one automated processor is further configured to rank a plurality of documents of the corpus of documents in dependence on the information and the computer-implemented generative model.

16. The method according to claim 13, wherein the multi-level transitive citation structure comprises at least one of citations, authors, cross references, and hyperlinks.

17. The system according to claim 13, wherein the computer-implemented generative model is a Bernoulli Process Topic model.

18. The system according to claim 13, wherein the latent topics are distinguished by Bayesian inference.

19. The system according to claim 13, further comprising an input port configured to receive a query, wherein the at least one automated processor is further configured to generate a search output responsive to the query selectively dependent on the computer-implemented generative model.

20. The system according to claim 13, wherein the at least one automated processor is further configured to generate a visualized output dependent on the at least one of representing, characterizing, clustering, summarizing, indexing, ranking, and searching of the documents in the corpus of documents.