RNN Hardware Implementation And Natural Language Processing Based On FPGA

Natural Language Processing(NLP)is a science that studies artificial intelligence to understand human language.It covers many fields,such as machine translation,text generation and speech recognition.As an important research direction in the field of artificial intelligence,Natural Language Processing has a good development prospect in the era of artificial intelligence.With the rapid development of Recurrent Neural Network(RNN),the research of Natural Language Processing has also made considerable progress.Recurrent neural network is usually implemented by software,using C,C++,Python and other languages to write algorithms,and then the central processor(CPU)completes the operation,but the software performs serial work in sequence,and can not reflect the parallel features and advantages of the neural network,and can not be transplanted to other hardware platforms,so that the application of Natural Language Processing has been limited.Field Programmable Gate Array(FPGA)has become an ideal device for the implementation of recurrent neural network with its high parallel computing,rich hardware resources,flexible logic unit and easy to develop chip.In this paper,the hardware implementation of Recurrent Neural Network based on FPGA,is proposed.Through the analysis of Recurrent Neural Network model,the hardware model of Recurrent Neural Network is designed and implemented.Based on that,the Tang poetry training set is used to train the Recurrent Neural Network,which verifies the feasibility of implementing Recurrent Neural Network and Natural Language Processing based on FPGA.At the same time,Recurrent Neural Network software implementation scheme based on tensorflow is designed.The two schemes are compared to verify the speed and power consumption advantages of an FPGA-based Recurrent Neural Network hardware implementation.