Design And Implementation Of Deep Learning Based Area Term Recogonition System

With the rapid development of mobile Internet technology and information technology,traditional paper documents are gradually being replaced by electronic documents.This progress has greatly reduced the time and effort that researchers wasted on investing and obtaining technical resources.Area term recognition is a task of discovering professional vocabulary of a specific field from related corpus of texts,which is of great significance for the semantic analysis of electronic documents and the rapid retrieval of text from corpora.Current Area term recognition process relies mainly on the combination of statistical model and linguistic rules to complete the process of term extraction and recognition.Efficient recognition process often come from a combination of several simple term recognition algorithms.The current terminology recognition technology relies mainly on traditional statistical methods.Some researchers have begun to explore the application of deep neural networks in the terminology recognition process,but have not yet formed a complete operational system.Therefore,based on natural language processing technology,this paper designs and implements a scalable and flexible implementation of area term recognition system.This paper mainly includes the following aspects:(1)Preprocessing the original text based on the N-gram model,and splitting the complete article into small text fragments.(2)A Bidirectional long-short term memory neural network model based on attention mechanism is proposed to extract terms from the text segment,and the conditional random field model is introduced to improve the recognition accuracy of long words.(3)Based on information entropy,a term credibility calculation method is proposed to improve the recognition accuracy.(4)The word vector model and the support vector machine(SVM)are introduced to automatically classify the extracted terms.As the experimental results shows,the system can effectively extract area terms from the original text and classify results according to the original term.The system is capable of reducing manual operations in the process of constructing term base.