A Research On Disease Knowledge Summarization Based On Semantic Relation And Link Analysis

Biomedical literature is growing rapidly in recent decades. The continued growth of biomedical literature brings great challenges to traditional information searching from texts. Information retrieval and information extraction greatly promote the transformation of traditional biomedical literature research. Biomedical information retrieval provides a convenient way which can return the retrieved documents to the user according to the query subject, such as online biomedical repository PubMed. However, the retrieval needs an appropriate query and returns large, unmanageable lists of citations, which makes it difficult for users to find the salient information they need. Information extraction solves the problem of reading a vast amount of literature on a particular topic, and provides text summarization. In this thesis, biomedical information extraction is used to extract biomedical literature knowledge based on semantic network, and giving a network visualization to display the extraction biomedical knowledge.This thesis extracts the salient semantic information of a seed disease related genes, drugs. Based on salient informaiton summarizaiton, a disease knowledge extraction visualization system is completed.This thesis proposes a deep knowledge extraction based on link analysis. A new summarization algorithm KM is used to identify the salient SemRep output. The relations most relevant to the seed topic are selected by the summarization algorithm. The hidden relations are extracted using deep search based on DFS from the directed unweighted graph of biomedical entities. The weak relations are filtered out by RRW and the final results are visualized. This thesis applies deep knowledge summarization on three diseases datasets. The experimental results show that the method is effective in terms of disease knowledge extraction.Biomedical knowledge summarization can effectively extract the important relations for each specific biomedical subject. Our method is much better than Combo algorithm, and is verified in multiple diseases corpus. Disease knowledge summarization can effectively improve the search efficiency of biomedical literature.