Research On Named Entity Recognition And Relation Extraction Between Entities Based On Depth Learning

With the rapidly development of the Internet and the explosive growth of network information,the research of information extraction technology has become more and more important.It has become a necessary requirement for the development of multiple disciplines and practical applications in many fields.Named entity recognition and entity relationship extraction are the most basic and important parts of information extraction technology research,because they are the basis of many complex tasks in natural language processing(such as knowledge map construction,machine translation,question answering system,and so on).In this paper we focus on named entity recognition and entity relationship extraction in information extraction.At present,among the methods of named entity recognition,the bidirectional LSTM combined with CRF method has achieved a good result,but there is still something to be improved.In this paper,we proposed a Generative Adversarial Nets suitable for the task of named entity recognition named Conditional Wasserstein Generative Adversarial Nets(CWGAN),inspired from Conditional GAN and improved Wasserstein GAN.Relative to the image probability distribution conditioned on textual description in CGAN,Our CWGAN obtains the NER label sequence probability distribution conditioned on sentence sequences.Both the generator and the discriminator use a bidirectional LSTM network.The difference is that the generator generates the probability distribution of the named entity tags,and the discriminator scores the generation quality of the generator and feeds it back to the generator.The generator updates the gradient according to the feedback to improve the quality of the probability of generating tags.In addition,we use gradient penalty in improved Wasserstein GAN to ensure that the gradient remains stable during backward propagation.Experiments show that the CWGAN model we proposed is effective in the task of named entity recognition.In the process of adversarial learning,the discriminator can guide the generator to be further improved in performance.In addition,the performance on accuracy of distant supervision combined withneural network model is satisfactory among numerous methods of entity relation extraction.However,there is often a lot of noise data in the labeled dataset obtained by distant supervision,which has a serous impact on the model of entity relation extraction.In this paper,we proposed a model of entity relation extraction via convolutional neural networks with improved attention mechanism.In this model,we find all positive instances which indicate the label relation as much as possible from the set of sentences containing the same entity pair.Then we use these positive instances to construct the combined sentence feature vectors,abandoning the possible noise sentences,so that we can minimize the impact of noise sentences and make full use of semantic information of all the positive instances.Experiments show that,the model we proposed is superior to the previous neural network relational extraction model in terms of both precision and recall.