Disease Related Knowledge Summarization Based on Deep Graph Search.

Area of Science:

• Biomedical Informatics
• Natural Language Processing
• Knowledge Discovery

Background:

• The rapid growth of biomedical literature presents challenges for traditional information retrieval (IR).
• Existing IR methods often yield unmanageable citation lists, failing to meet specific information needs.
• Effective summarization of disease-related knowledge is crucial for researchers.

Purpose of the Study:

• To develop an automated approach for disease-related knowledge summarization from biomedical literature.
• To improve the extraction of salient disease information and uncover indirect relationships between biomedical entities.
• To enhance the precision and recall of information retrieval for disease knowledge.

Main Methods:

• Utilized Kullback-Leibler Divergence and mutual information for salient information extraction.
• Employed a deep search strategy based on depth-first search (DFS) to identify hidden biomedical entity relations.
• Applied a random walk algorithm to filter weak or irrelevant relations.

Main Results:

• The proposed approach achieved 60% precision and 61% recall for salient information extraction concerning Carcinoma of the bladder.
• Demonstrated superior performance compared to the Combo method in experimental evaluations.
• Successfully extracted key disease information and identified indirect relationships.

Conclusions:

• The developed method offers an effective solution for automatic disease knowledge summarization.
• This approach enhances the ability to discover and utilize complex information within the biomedical literature.
• The findings suggest a significant improvement in information retrieval accuracy for disease-specific research.