Research On Entity Relationship Extraction Based On Deep Learning

As an important basic work in the field of natural language processing,entity relationship extraction is a key part of the construction of knowledge graphs.Its purpose is to discover entities in texts and determine the semantic relationships between entities.This research is helpful to the construction of knowledge bases.And then provide powerful services and support for smart search,smart recommendation,smart question and answer,etc.Early methods rely heavily on researchers to artificially design specific templates or artificially design related features according to the characteristics of the corpus.This process requires a high cost of time and is difficult to promote and apply in various fields.Due to the continuous breakthroughs and development of deep learning technology in recent years,researchers have begun to study the use of deep learning to solve the task of entity relationship extraction and continue to make new breakthroughs in this task.However,these methods still have problems such as dependence on domain knowledge base features,insufficient feature learning capabilities for data,and lack of further research on feature optimization and combination.This article focuses on the above three points,so that the new entity relationship extraction model proposed in this article can automatically mine the rich features contained in the data itself,without relying on various external knowledge base resources.The research content and innovations of this paper are as follows:(1)Propose a relationship extraction model for fusion of word features and features between adjacent wordsThis paper proposes a relationship extraction model that combines word features and features between adjacent words.This model can make full use of the characteristics of convolutional neural networks and bidirectional long-term memory networks,and combine the attention mechanism to extract word-based sentence features.The sentence characteristics of the relationship between adjacent words can further dig out the semantic information latent in the natural language text.The experimental results show that the experiments designed in this paper are trained and tested on the public Sem Eval-2010Task8 and Wiki80 data sets,and the test results are improved compared to the other three classic and mainstream models.(2)Propose a relationship extraction model that combines multiple feature combinations with entity-related informationThe entity and the contextual information around the entity help determine the semantic relationship of the entity in the sentence,which plays a very important role in the task of relationship extraction.Therefore,without relying on external knowledge,in order to enable the model to make better use of entity information and learn important context vector information related to the entity in the sentence,this paper constructs a sentence that can learn the continuous features of the sentence and capture A neural network model of the complex background relationship between entities and entities in sentences.Finally,this article optimizes the combination of various features learned by the model.The experimental results show that the multi-feature combination relationship extraction model proposed in this paper,which combines entity-related information,effectively improves the entity relationship extraction effect and is better than the comparison model.