The Entity Recognition And Application Of Human Relationship Analysis Based On Deep Learning

With the development of the Internet,text stores more and more information,so how to extract useful information from these unstructured text information is very important.The transformation of unstructured text into structured text is the most basic step in information utilization,while the key of structured text lies in named entity recognition and relationship extraction,which is the most basic task of natural language processing.Named entity recognition and relationship extraction play an indispensable role in text understanding,information retrieval,text summarization,question answering,machine translation,and knowledge base building.In this paper,named entity recognition is studied from coarsegrained entity recognition and fine-grained entity recognition.At the same time,the named entity recognition model is used to extract the data set of names and relationships between people,and the model of relationships between people is constructed.The specific contents are as follows:1)Puting forward a kind of named entity recognition method based on BCNN-BiLSTM,this method is introduced in BERT training models of the convolutional neural network combined with deep learning network and Long Short-Term Memory network in the field of open data entity recognition,using Bi-LSTM for character information,convolution neural network for sentence information.The obtained information matrix was spliced,and the conditional random field(CRF)was used to annotate the sequence to complete the extraction of the names of people,places and organizations.2)A fine-grained named entity recognition method based on attention mechanism is proposed.First,the information of fine-grained entity data is captured using the Long ShortTerm Memory network.Then,the attention mechanism is used to set different weight parameters for different location information to distinguish the importance of different location information and improve the accuracy of fine-grained entity recognition.Finally,conditional random field is used to realize fine-grained entity recognition.3)A character relationship extraction method based on deep learning is proposed.In this method,the data were vectorized and the features were acquired through Long ShortTerm Memory network,residual network and channel attention.During the acquisition process,the multi-layer convolution of residual network is used to reduce the loss of matrix information,and the important information and important features of character relation words are extracted dynamically by channel attention,which improves the extraction effect and builds a more effective character extraction model.The experimental results show that the proposed method is effective.In the BCNN-BiLSTM-based entity recognition model,the F1 values of People's Daily corpus and Weibo corpus reach 91.3% and 66.3% respectively,which can complete the extraction of names of people,places and organizations.The fine-grained entity recognition model based on attention mechanism can also be used to extract many kinds of entities such as scenic spots and movie names.In the character relationship extraction model,the F1 value reaches70.51%.The character name and character relationship can be extracted,and the F1 value reaches 85.87%,which has a certain improvement compared with other methods.