Intention recognition method and device and storage medium

Abstract

The embodiment of the invention discloses an intention recognition method and device and a storage medium. The method comprises the steps of obtaining a current voice and a previous answer of a user;wherein the previous answer represents an answer having a preset association degree with the current dialect in at least one answer corresponding to at least one voice before the current voice; processing the current voice and the previous answer based on a graph neural network model to obtain features corresponding to the current voice and the previous answer; performing preset times of processing on the features corresponding to the current speaking and the previous answer based on an attention model, and extracting features of preset dimensions in the features corresponding to the current speaking and the previous answer; and determining an intention corresponding to the current speaking of the user based on a feature of a preset dimension in the features corresponding to the current speaking and the previous answer.

Description

technical field [0001] The present invention relates to the technical field of intelligent dialogue, in particular to an intent recognition method, device and storage medium. Background technique [0002] Computers play an indispensable role in all aspects of modern society. Natural language intelligent recognition with computer intelligent recognition application technology as the core is developing rapidly and slowly entering the sight of our public. A dialog system based on computer intelligent recognition application technology is generally divided into two parts: intent recognition and dialog management. Intent recognition is used to identify the user's purpose or task to be completed, and dialog management is responsible for giving answers and filling slots. The reason why dialog-based intent recognition is a challenging task is that the user input in the dialog is usually very short and lacks enough information, requiring the dialog system to accurately understand the...

AI Technical Summary

Problems solved by technology

[0003] At present, the methods of intent recognition are generally concentrated in the context of a single-round dialogue. On the basis of word segmentation of user message texts, intent recognition is regarded as a classification problem, and then machine learning models are used for supervised training. Vectorization often cannot be combined with specific tasks, so that the high degree of fit between the word vector and the task cannot be guaranteed, which may cause the model to converge slowly or have a greater probability of falling into a local optimum, and the cost of pre-training is generally large; even if there are multiple rounds The context of the dialogue also directly introduces the overall coding of the dialogue text before the current dialogue into the current analysis. Generally, the identity of the user and the respondent is not distinguished, and the previous text is generally introduced into the current analysis as a whole, which makes the current answer more accurate for the user's next message The role of guidance is ignored. In fact, the answer of each step has a strong correlation with the user's next message.

Images