| With the development of network technology,more and more academic literature is published in the form of electronic documents.Biology,as an experimental discipline,the biological literature mostly discusses experimental findings,such as regulation,function and expression information about gene,phenotype and protein.With the increasing attention to life sciences,information retrieval and understanding in the biological literature has received more and more attention.Image and table information often shows the core information such as experimental process and experimental results in a document,so the interpretation of the table and image information in the literature is especially important.How to extract and interpret these results through text mining from a lot of literature has important research significance.This thesis studies the table mining in the biological literature.Since the literature in XML format covers only a small amount of literature,this thesis uses the literature in PDF format.At present,the general table mining tools are not work well in the biological literature,so this thesis proposes a method for extracting and semantically understanding the tables of biological literature.In this thesis,the QTL table is taken as the research object,and the QTL information extracted from the literature is collected and interpreted.The main research work is as follows:(1)A table extraction process for biological literature is proposed,including table detection,structural analysis,and information extraction.As the table in the biological literature is in the form of a three-line table,based on the binary image,this thesis use the connected value to detect table and restore the structure of table.Due to the large differences in the header structure of different authors,we standardized the headers of heterogeneous tables.According to experiment result,we can get better F values than the table mining tool TableSeer and document conversion tool BCL.(2)A method for semantic analysis of tables in the literature based on contextaware is established.First,the extracted tables are filtered based on the semantic triples,and the table columns are filtered based on predefined rules.Then the table is semantically interpreted from two aspects: semantics of table data and table related text.For the table data,the external knowledge base is used to semantically annotate biological entities in the table.For text data in the literature,the texts related to the table are filtered to form a summary.Finally,the dependency parse tree is used to analysis summary sentences.Based on the rules,extracting supplementary information such as methods,places,and parents from the sentences.(3)A QTL database containing nine crops was constructed.QTL data is widely concerned by biological researchers as data on associated quantitative traits and genes.Literatures about QTL mapping were searched and downloaded from the PubMed database.Based on the above table mining process,the QTL tables are mined from the literature.In order to get further biological interpretation of the extracted QTL data,combined with the reference genomic information and gene information of the species,the QTL interval is mapped from the genetic map to the physical map.The genes in the QTL interval were obtained by physical map and QTL data is visualized in the map,which will provide convenience for research of the traits and genes. |
