Knowledge Graph Construction Of Domain Literature

The domain literatures are the most frequently used materials by researchers in their indepth studying process of a specific subject area.In a big-data era of scientific research,the traditional search methods are unavailable to provide the accurate understanding of semantic information and the quick locating of the needed knowledge for researchers facing massive domain literatures.However,the knowledge graph emerging as a new information technology can visually present the relationship between the knowledge entities and other entities from the perspective of semantic.Applying the knowledge graph to professional literatures in scientific research to form the knowledge graph on professional literatures is thus of great application value for researchers and of convenience for retrieving professional literatures.The main research results are as follows.1.This thesis uses the Chinese literatures on the crop hyperspectral domain as the metadata,and adopting the top-down construction method,using keywords as dictionaries and article titles as auxiliary abstract markers,and by these,a corpus on the crop hyperspectral domain is constructed,three types of entities and three types of relationships are put forward on statistical classification,and then a knowledge graph is established against the unstructured data like the abstracts of literatures.2.Two deep learning models,the BERT-Bi-GRU-CRF and the BERT-Bi-LSTM-CRF,are comparatively used in recognizing the named entities of a knowledge graph,and the applicability of the BERT-Bi-GRU-CRF in this domain is experimentally verified.In this thesis,the relationship extraction of literature entities in crop hyperspectral field based on the Bi-LSTM-Attention model is carried out to complete the construction of knowledge map in crop hyperspectral field.3.The constructed knowledge graph of the crop hyperspectral domain literatures is stored in the Neo4 j graph database,on which an application system of literature knowledge graph is designed and implemented,and through the system the querying and visual displaying of knowledge entities is performed,which shows a reference for the upper application of a knowledge graph.By studying the construction of a knowledge graph of literatures in the crop hyperspectral domain,the thesis sets up an application system of knowledge graph on literature for researchers in the domain,which is helpful for researchers to improve the efficiency of their literature reading and provide a reference for constructing knowledge graphs in other domains.