NLP-QA Framework Based on LSTM-RNN

Abstract

The paper mainly majors in question answering system(QA). The original natural response theory originated from Alan Turing's Turing Machine Theory in 1950. Nowadays, the best way to implement a QA system is the Deep Learning. This project is based on the Seq2Seq model theory, I design and implement an automatic question answering system model based on LSTM-RNN algorithm. The paper completely describes the framework and design ideas of the entire system. it realizes the following aspects:1)A Seq2Seq model based on LSTM-RNN.2) We design the QA framework adapts to different chat scene. Results show that the average of perplexity is 2.92. And model's loss is 1.07. To some extent, it reduces the resistance of deep learning in developing. The design and implementation of the framework simplifies the development process and facilitates the user's own data set.