Statistical Machine Translation Algorithm Based on Improved Neural Network

Abstract

There are rules-based machine translation and modulate-based machine translation but they are all based on complex and hardly-summarizing language rules in essence. This paper discusses necessity and possibility of combination between NN(neural network) and traditional search methods, points advantages and disadvantages of NNMT(neural network machine translation) and puts forward a new MT intelligence integration system framework. It can partially solve some contradictions. If it effectively fuses multi-channel to acquire knowledge such as traditional rules acquisition method, NN method, KDK(knowledge discovery in knowledge base) and KDD(knowledge discovery in database), which largely enhances system solution and relieves bottleneck of grammatical semantic rules acquisition to improve overall performance of machine translation.