Research On Entity Relation Extraction From Chinese Text In Computer Field

Information extraction can automatically extract knowledge from massive data and express it in a structured way,which greatly reduces the workload of manual data integration and knowledge extraction.With the rise and rapid development of artificial intelligence technology,natural language processing technologies such as text classification,Knowledge Q & A and machine translation have made new breakthroughs in many technical difficulties.However,as a basic research,information extraction is widely used and plays an important role in many natural language processing technologies.As an important sub task of information extraction,the effect of entity relationship extraction will determine the quality of information extraction.With the rapid development of machine learning,neural networks,one of the main machine learning methods,such as convolutional neural network CNN,cyclic neural network RNN and long-term and short-term memory network LSTM,are also widely used in relationship extraction tasks because they reduce the requirements for feature engineering ability and have the advantage of super learning ability.However,these e-learning focuses are different,and there are still some shortcomings and limitations.In natural language processing tasks such as relationship extraction,the CNN based model focuses more on capturing the local features of the input text,but it can not deal with the long-distance relationship between words in the text.On the contrary,the model based on LSTM has obvious advantages in long-distance sequence feature extraction,but it is not as good as CNN in local feature extraction.In addition,in the task of relationship extraction,the existing Chinese data sets are small,and there are still problems such as annotation quality.Therefore,most of the research on relationship extraction is mainly carried out on English data sets,and it is also of great research value and significance to build high-quality Chinese data sets and carry out relationship extraction related work.The main contents of thesis are as follows:1.Collect data sets in the computer field,conduct strict manual review and labeling,and construct a small-scale Chinese data set for relationship extraction.2.A Bi LSTM-CNN-Attention entity relationship extraction model is built,trained and tested on the self-built data set,so that the model reaches a high performance state.3.Different word vectors can represent different semantics of words to a certain extent,so as to overcome the problem of word meaning omission caused by the single representation of word vectors.This paper proposes a method of splicing word vectors,so as to enrich the semantic information contained in word vectors.The experimental results show that the proposed method has obvious advantages over the previous methods.