Research On The Construction Method Of Knowledge Map In Ancient Chinese

The rapid developing natural language processing technology has been more and more frequently used in life.People use this technology for interest recommendation,machine translation and making question answering system.Nowadays natural language processing applications are concentrated on English or modern Chinese,but the application of ancient Chinese is still relatively small.Why the research in ancient Chinese is not enough is that there are some differences between ancient Chinese and modern Chinese.Most ancient Chinese texts are unstructured,there is a problem that the same person would use different names in the same article.How to transform unstructured ancient Chinese text and solve the problem of one person with different names is the focus in this paper.At the same time,I also proposed a new neural network model to extract the relationship between entities,which can be used to expand the knowledge in the ancient Chinese knowledge map.What the research focuses on of this article is processing of ancient Chinese data and constructing a knowledge map based on the obtained data,the corpus selected for the subject is Chinese classical cultural article.The ancient Chinese text belongs to unstructured data,which needs to be cleaned first.The processed data is stored in the graph database and realizes query or other functions.We use a novel as the original corpus.First,each sentence in the novel is segmented,and we identify named entities in the sentence which has been segmented.Then,a third-party knowledge base is used to label the data after the identified entity to obtain structured data.The data set is composed of entities,relationships between entities,and the context in which the entities exist,we removed some of the noise data introduced.An algorithm is proposed in this experiment to solve the problem of multiple names in structured text.Another work in this paper is that we propose a new type of neural network structure BLSTM-CNN used in the relationship extraction task.The idea of this model mainly comes from the bidirectional long-short-term memory network and the convolutional neural network.It has excellent performance in the classification of entity relationships.This model gets a very good result on the classic relationship extraction task SemEval2010_task8,it is better than the attention-based bidirectional long short-term memory networks which is widely used today with the best results.Then this neural network model is applied to the ancient Chinese data set for the relationship prediction task.Finally,we extract the triples in the structured data to establish a knowledge graph,and visually display the relationship between entities ancient Chinese.