Research On Entity Relation Extraction Based On Deep Learning

Entity relation extraction can automatically extract knowledge from the massive data of the internet to form structured data,reducing the workload of manually collating data and extracting knowledge,which is of great significance.Traditional entity relation extraction methods mostly rely on natural language processing tools,so the performance of entity relation extraction depends heavily on the level of natural language processing tools.The errors caused by the use of natural language processing tools may also interfere with the entity relation extraction process,resulting in degraded model extraction performance.The method of entity relation extraction based on deep learning is studied.The text is processed without natural language processing tools.Before the relation is extracted,the distributed word embedding vector is used to represent the text,and the words are embedded into the low-dimensional space to preserve the context.In addition to the use of convolutional neural networks and recurrent neural networks for the study of entity relation extraction respectively,two types of neural networks are combined to study the entity relation extraction.Firstly,the convolutional neural network and the recurrent neural network are used to study the entity relation extraction.The word vector and position features,local features and sequence features in the text are studied.The effects of these features on the performance of the entity relation extraction model are analyzed.Then,the convolutional neural network and the recurrent neural network are combined,and the combined model of the series and parallel is used to study the entity relation extraction.Without natural language processing tools,the combined model obtains the extraction effect close to the traditional machine learning using the natural language processing tools.Finally,the multi-model joint extraction method is adopted.The experimental results show that the extraction method achieves better performance than the traditional machine learning method.