Context attention stream-based dialogue intention recognition system and method

Abstract

The invention provides a context attention stream-based dialogue intention recognition system and method. The system comprises an input coding module, an autocorrelation coefficient analysis module, a feedforward neural network and a multi-task learning module. The input coding module is used for coding an input statement containing a plurality of words to obtain a corresponding representation vector; the autocorrelation coefficient analysis module is used for performing calculation after splicing the representation vector of the current statement and the representation vector of each historical dialogue statement to obtain a previous statement representation vector fused with question information, performing feature fusion according to the previous sentence representation vector to obtain a context sentence representation vector fused with dialogue context information, and finally, performing dot product operation according to the representation vector of the current statement and the context sentence representation vector to obtain a feature vector for intention recognition. The multi-task learning module is used for optimizing the feature vector according to the total loss function of the system, so that the efficiency and accuracy of dialogue intention recognition are improved.

Description

technical field [0001] The present invention relates to the field of computer technology, in particular to a dialogue intention recognition system and method based on context attention flow. Background technique [0002] The core functional module of the dialogue robot is intent recognition. The robot first needs to predict the corresponding intent based on the dialogue sentence sent by the user, and then send the corresponding answer to the user based on the intent, thereby completing the online automatic answer. Dialogue is a multi-round question-and-answer process. However, when online robots recognize intent, they only consider the content of a single sentence, and the intent in many conversations cannot be identified through the content of a single sentence. Therefore, the online robot has quite a few sentences whose intentions cannot be accurately identified based on a single sentence, which leads to the failure of the robot's question-and-answer response. [0003] I...

AI Technical Summary

Problems solved by technology

The model based on the memory network cannot directly select relevant historical dialogue sentences as the above-mentioned dependent information of the dialogue, which makes it difficult for the model to accurately integrate the above-mentioned dialogue information into the current sentence

In addition, the encoding features of some of the sentences may be repeatedly selected each time, causing the model to fail to fully pay attention to other related features, affecting the ability of the model to model multiple rounds of dialogue

Images