Natural Language Processing Based on Convolutional Neural Network and Semi Supervised Algorithm in Deep Learning

Natural language (neural network) is composed of a large number of human visual systems. Its research began with artificial perception and is used to describe things or phenomena involved in the process of human brain processing external information. At the same time, neural network is a nonlinear system description and analysis method. It is mainly used in manual calculation, pattern recognition and so on. Therefore, this paper studies natural language processing based on convolutional neural network and semi supervised algorithm in deep learning. Firstly, this paper expounds the definition of natural language processing, then studies the application of convolutional neural network and semi supervised algorithm, and tests the feature recognition of word vector in natural language to show the optimization results of deep learning algorithm. The test results show that the effect of word vector merging is relatively high. In the optimization process of convolutional neural network, after adding standard interval for semi supervised learning, the accuracy of prediction results also increases significantly, and with the increase of the number of labels, the accuracy also increases gradually.

I Introduction

With the rapid development of computer and Internet, human beings have entered the new information age. Natural language processing plays an irreplaceable role in the information age. The emergence of natural language processing as a research direction mainly solves the problem of enabling computers to understand human natural language. Natural language understanding completes the interaction between human and machine, and makes the corresponding command control processing work required by human. This shows the importance of natural language processing research. The development of natural languages (such as pronunciation, grammar, etc.) has been a long and slow process. As time goes by, it can no longer meet people’s demand for information retrieval and understanding, and is gradually applied to various fields by artificial neural network. In natural language processing, it is a difficult task to implement question analysis in Chinese question answering system because of the great differences between Chinese and western languages in sentence patterns, semantics and words. In the problem analysis stage, word segmentation and problem classification are the key steps in the whole problem analysis stage. The results of the processing will directly affect the understanding degree of the subsequent steps to the problem analysis, and ultimately affect the accuracy of correctly analyzing the real intention of users. After the problem statement is preprocessed, the final problem classification should be done in the problem classification link. The efficient word segmentation and question classification method can not only greatly reduce the search scope and shorten the retrieval time, but also provide accurate direction for the subsequent answer extraction stage with its accurate word segmentation and classification effect.