A Neural Network Model for Chinese Sentence Generation with Key Word

Using a word to make a sentence is a basic skill for human. It reflects the abilities of creativity and the degree of mastery of language skills of human. If the machine has this ability, which is used in nature language processing, it will be beneficial to solve some sentence generation problems, specially for summarization generation, machine translate, automatic question and answer and other application scenarios. Long Short-Term Memory(LSTM), is a neural network model broadly used in data processing and predictions. But in the field of text generation, there still are many limits. Most text generation based on LSTM, can only generate word on one direction from start word; or it can not keep the key word in the result. Our paper is to explore a way for LSTM to generate the context before and after the key word. This paper will provide a kind of LSTM model to achieve the function, generating sentence from key word and keeping the key word in the result sentence for Chinese.