REQUEST REPHRASING SYSTEM, METHOD FOR TRAINING OF REQUEST REPHRASING MODEL AND OF REQUEST DETERMINATION MODEL, AND CONVERSATION SYSTEM

摘要

Provided is a request rephrasing system 120 with which a conversation system can flexibly respond to various requests having different manners of expression, said request rephrasing system 120 comprising: a pre-processing unit 130 for converting a user input 56 into a word vector sequence; and a neural rephrasing model 94, pre-learned by machine learning, for receiving the word vector sequence as input and rephrasing the request represented by the word vector sequence as a request which a question/response device 122 has a higher probability of being able to respond to than before the rephrasing thereof. The system may involve determining whether the user input 56 is a request for the preprocessing and carrying out the rephrasing of the request only if the user input 56 is determined to be a request. In addition, the system may involve classifying which request classification the inputted request belongs to using a classification model 98 and inputting said classification into the neural rephrasing model 94 as one feature in the rephrasing.