Research And Implementation Of Medical Literature Based Discovery System Based On Linked Data

There is a great deal of medical knowledge in medical literature. It is of great significance and value to carry out knowledge discovery. In the process of medical literature based discovery, on the one hand, due to the characteristics of update frequency of dictionary, traditional literature based discovery model can not keep up with the development of biomedical research, resulting in relatively new words can not be extracted, on the other hand, the biomedical entity itself has its special semantic relations, need to match relations according to the external knowledge base. In this paper, we consider the deficiency of traditional knowledge discovery model in the field of medicine, and study the knowledge discovery model based on the linked data.This paper presents the medical literature based discovery model based on the linked data (MLBDM), the model constructed an entity database of medical relations based on linked data, used the combined method of statistics and dictionary to complete the named entity recognition in the medical literature, and associated corresponding data with the medical entity in the database. After matching the semantic relations of the entities in the starting literature collection, the medical entity relation graph was constructed, the entities with the highest weight are selected by the graph ranking algorithm, and the characteristic words of these entities are retrieved as the intermediate keywords. After the collection of documents, repeated the extraction steps of medical entities in the intermediary literature collection, and according to the sorting algorithm, selected the highest weight entity, described it as the knowledge of discovery.Based on the knowledge discovery model, this paper designed and realized the knowledge discovery system of medical literature. Firstly,DBpedia and SNOMED CT were used as the data source. The named entity recognition method based on conditional random field and TRIE tree was used and the method of selecting the entities used PageRank as the graph ranking algorithm. The system acquired the starting medical literature collection from the PubMed literature database according to the keyword input by the user and used it as the model input for knowledge discovery. This paper demonstrates the effectiveness of the model and the availability of the system by demonstrating the trail of Professor Swanson's classic fish oil and Raynaud's disease.