Research On Entity Relation Extraction Based On Convolutional Neural Network

With the rapid development of information technology,the Internet has produced a large number of unstructured text data,including news,blog,government documents,chat logs,etc.How to help people get valuable information from the web quickly becomes the problem concerned by the scholars of computer science areas.Entity Relation Extraction is generated in this context.Its main task is to identify the entities in the text and extract the semantic relations between the entities.The existing method based on supervised learning has achieved good results,but they rely heavily on POS,syntactic parsing,and other natural language processing tools to provide classification features.And these tools tend to have a lot of errors.These errors will continue to propagate in the relationship extraction system.In order to avoid relying on complex feature engineering,this paper proposes to use the convolution neural network to solve the problem.It will automatically learn features from the sentence to construct the sentence's Distributed representation as the input of the relational classification model.Thereby minimizing the reliance on some NLP processing tools and resources.And design and build a company relationship extraction system for Internet news text.The main work of this paper is as follow:1.In view of the existence of the company entity relation showed in the Internet News,this paper proposes a method of constructing relational corpus based on Bootstrapping technology.This method overcomes the shortcomings of time consuming and laborious in the process of manual labeling.2.In the case of the shortcoming of the traditional word bag model lacks semantic information and the location information of the word when representing sentences,this paper propose two methods based on weighting of Word Embedding and Convolutional Neural Network to construct a compact and semantic Distributed representation of the sentences.As the input of the relational classification model.3.On the basis of the first part of the work,combining the key technology of web page content extraction and Named Entity Recognition,this paper constructs the prototype of enterprise relationship extraction system.