SYSTEM AND METHOD FOR AUTOMATICALLY GENERATING SYSTEMATIC REVIEWS OF A SCIENTIFIC FIELD

• US 20140195539A1
• Filed: 10/01/2013
• Published: 07/10/2014
• Est. Priority Date: 05/28/2010
• Status: Active Grant

• Current Assignee: Drexel University
• Original Assignee: Drexel University
• Inventors: Chen, Chaomei
• Granted Patent: US 8,935,291 B2
• Time in Patent Office: Days
• US Class Current: 707/739
• CPC Class Codes: G06F 16/285 Clustering or classification
• CPC Class Codes: G06F 16/367 Ontology

Abstract

A system and method are provided for automatically generating systematic reviews of received information in a field of science and technology, such as scientific literature, where the systematic review includes a systematic review of a research field in the scientific literature. The method includes the steps of constructing a time series networks of words, passages, documents, and citations and/or co-citations within received information into a synthesized network, decomposing the networks into clusters of fields or topics, performing part-of-speech tagging of text within the received information to provide tagged text, constructing semantic structures of concepts and/or assertions extracted from the source text, generating citation-based and content-based summaries of the clusters of fields or topics and the semantic structures, and generating structured narratives of the clusters of fields or topics and the summaries of the generated semantic structures. Narratives of the citation-based and content-based summaries are merged into a systematic review.

Claims

1. A computer-implemented method for automatically generating systematic reviews of information received from a source text in a field of literature, comprising a programmed processor performing the following steps:

constructing associative networks of entities within the received information;

decomposing the associative networks into clusters of fields or topics;

performing part-of-speech tagging of text within the received information;

constructing semantic and ontological structures and/or assertions extracted from the source text;

generating citation-based and content-based summaries of the clusters of topics or fields and the semantic and ontological structures; and

generating structured narratives of the clusters of field or topic-characterizing entities and the summaries of the generated semantic structures.

2. The method of claim 1, further comprising the programmed processor performing the step of merging narratives of the citation-based and content-based summaries into a systematic review having a predetermined arrangement.

3. The method of claim 2, wherein generating citation-based and content-based summaries of the clusters of fields or topics and the semantic structures comprises the processor measuring the saliency, novelty, significance, and transformative features of individual entities in the clusters of fields or topics and semantic structures.

4. The method of claim 3, wherein generating structured narratives of the cluster of fields or topics and the summaries of the generated semantic structures comprises the processor labeling and summarizing features of the clusters of fields or topics and delinearizing the characteristics of such clusters into templates that provide summarizations of the structure and trends of the topic or field evolution in the clusters of fields or topics at multiple levels of abstraction.

5. The method of claim 1, wherein the field of literature comprises scientific literature and the systematic review comprises a systematic review of a research field in the scientific literature.

6. The method of claim 1, wherein constructing associative networks of publications including semantic networks within the received information comprises the processor selecting node types and link types for each time slice of the received information, computing similarity or proximity scores for the nodes, constructing networks of the node information, and merging respective networks from different time slices.

7. The method of claim 6, wherein decomposing the associative networks into clusters of topics or fields comprises the processor clustering nodes and measuring quality of the clustering by calculating modularity and/or mean silhouette scores for the clusters of fields or topics.

8. The method of claim 7, wherein generating citation-based and content-based summaries of the clusters of fields or topics and the semantic structures comprises the processor identifying citers to and cited members of a cluster, summarizing structural and temporal properties of the cluster, computing metrics of saliency and novelty for an associative network formed by the cluster, ranking the clusters based on the saliency and/or novelty metrics, and generating structured narratives from the ranked clusters.

9. The method of claim 1, wherein generating structured narratives of the clusters of fields or topics and the summaries of the generated semantic structures comprises the processor selecting a narrative template from a set of predefined templates.

10. The method of claim 1, wherein performing part-of-speech tagging comprises the processor annotating the received information by a type of each word in the received information.

11. The method of claim 10, wherein performing part-of-speech tagging further comprises the processor segmenting the received information into sentences, paragraphs, or other types of passages.

12. The method of claim 1, wherein constructing semantic structures of concepts and/or assertions extracted from the tagged text comprises the processor constructing a structured representation of concepts and a semantic network of assertions in the received information.

13. The method of claim 12, wherein constructing semantic structures of concepts and/or assertions extracted from the tagged text further comprises the processor merging a newly constructed semantic structure with an existing semantic structure to differentiate different sources for the newly constructed and existing semantic structures.

14. The method of claim 13, wherein constructing semantic structures of concepts and/or assertions extracted from the tagged text further comprises the processor ranking merged semantic structures based on saliency and novelty, generating narratives of top ranked concepts and/or assertions in the received information, and merging generated narratives in a predetermined order.

15. The method of claim 1, further comprising quantifying a potentially creative work by constructing quantitative metrics from structural variations in the work due to newly introduced concepts, conceptual relations, and conceptual structures where structural variations are computed as a degree of deviation from a network structure of contemporary knowledge.

16. A system for automatically generating systematic reviews of received information in a field of literature, comprising a processor programmed to perform the following steps:

constructing associative networks of entities within the received information;

decomposing the associative networks into clusters of topics or fields;

performing information extraction within the received information to provide tagged text;

constructing semantic structures of concepts and/or assertions extracted from the tagged text;

generating citation-based and content-based summaries of the clusters of topics or fields and the semantic and ontological structures; and

generating structured narratives of the clusters of topics or fields and the summaries of the generated semantic structures.

17. The system of claim 16, wherein the programmed processor further performs the step of merging narratives of the citation-based and content-based summaries into a systematic review having a predetermined arrangement.

18. The system of claim 17, wherein the programmed processor measures the saliency, novelty, significance, and transformative features of individual entities in the clusters of fields or topics and semantic structures.

19. The system of claim 18, wherein the programmed processor labels and summarizes fields or topics of the clusters of fields or topics and delinearizes the resulting clusters into templates that provide summarizations of the structure and trends of the information in the clusters of topics or features at multiple levels of abstraction.

20. The system of claim 16, wherein the field of literature comprises scientific literature and the systematic review comprises a systematic review of a research field in the scientific literature.

21. The system of claim 16, wherein the programmed processor constructs associative networks of entities within the received information by selecting node types and link types for each time slice of the received information, computing similarity or proximity scores for the nodes, constructing matrices of the node information, and merging respective matrices from different time slices.

22. The system of claim 21, wherein the programmed processor decomposes the associative networks into clusters of fields or topics by clustering nodes and measuring quality of the clustering by calculating modularity and mean silhouette scores for the clusters of fields or topics.

23. The system of claim 22, wherein the programmed processor generates citation-based and content-based summaries of the clusters of fields or topics and the semantic structures by identifying citers to and cited members of a cluster, summarizing structural and temporal properties of the cluster, computing metrics of saliency and novelty for an associative network formed by the cluster, ranking the clusters based on the saliency and/or novelty metrics, and generating structured narratives from the ranked clusters.

24. The system of claim 16, wherein the programmed processor generates structured narratives of the clusters of fields or topics and the summaries of the generated semantic structures by selecting a narrative template from a set of predefined templates.

25. The system of claim 16, wherein the programmed processor performs part-of-speech tagging by annotating the received information by a type of each word in the received information.

26. The system of claim 25, wherein the programmed processor performs part-of-speech tagging by segmenting the received information into sentences, paragraphs, or other types of passages.

27. The system of claim 16, wherein the programmed processor constructs semantic structures of concepts and/or assertions extracted from the tagged text by constructing a structured representation of concepts and a semantic network of assertions in the received information.

28. The system of claim 27, wherein the programmed processor constructs semantic structures of concepts and/or assertions extracted from the tagged text by merging a newly constructed semantic structure with an existing semantic structure to differentiate different sources for the newly constructed and existing semantic structures.

29. The system of claim 28, wherein the programmed processor constructs semantic structures of concepts and/or assertions extracted from the tagged text by ranking merged semantic structures based on saliency and novelty, generating narratives of top ranked concepts and/or assertions in the received information, and merging generated narratives in a predetermined order.

30. The system of claim 16, wherein the programmed processor further quantifies a potentially creative work by constructing quantitative metrics from structural variations in the work due to newly introduced concepts, conceptual relations, and conceptual structures where structural variations are computed as a degree of deviation from a network structure of contemporary knowledge.

31. A computer readable storage medium having instructions stored thereon that when processed by a processor cause the processor to implement a method for automatically generating systematic reviews of information received from source text in a field of literature, wherein the instructions, when executed by the processor, cause the processor to perform the following steps:

constructing associative networks of entities within the received information;

decomposing the associative networks into clusters of fields or topics;

performing part-of-speech tagging of text within the received information;

constructing semantic and ontological structures of concepts and/or assertions extracted from the source text;

generating citation-based and content-based summaries of the clusters of topics or fields and the semantic and ontological structures; and

generating structured narratives of the clusters of field or topic-characterizing entities and the summaries of the generated semantic structures.

32. The storage medium of claim 31, further comprising instructions that when executed by the processor cause the processor to perform the step of merging narratives of the citation-based and content-based summaries into a systematic review having a predetermined arrangement.

33. The storage medium of claim 31, further comprising instructions that when executed by the processor cause the processor to perform the step of quantifying a potentially creative work by constructing quantitative metrics from structural variations in the work due to newly introduced concepts, conceptual relations, and conceptual structures where structural variations are computed as a degree of deviation from a network structure of contemporary knowledge.