Intention recognition system and method

Abstract

The invention discloses an intention recognition system and method.The intention recognition system comprises a data processing module, a characterization module, a feature extraction module and a target function module, wherein the data processing module is used for processing an input text to obtain input data; the representation module is used for mapping input data to obtain a word vector, a word position vector and a sentence index vector of a word; the feature extraction module is used for analyzing the character vector, the character position vector and the index vector of the sentence where the character is located to obtain features required by model analysis; the objective function module sets an objective function suitable for the intention recognition system and is used for evaluating the intention recognition system. According to the method, an intention recognition system with very high accuracy can be trained by only needing a small amount of data, the manual labeling cost is greatly reduced, and the method is suitable for actual scenes in which the number of company labeling personnel is small or excessive labeling cannot be performed due to a large amount of data.

Description

technical field [0001] The present invention relates to the field of natural language processing, in particular to an intent recognition system and method. Background technique [0002] The development of the Internet and mobile Internet has provided convenience for users to shop online. In order to improve user experience, these e-commerce platforms, product platforms, and service platforms will provide online manual services, but the labor costs will be relatively high. Compared with the number of users On a large platform, manual services have not been able to meet the needs of buyers and users. When there is a bottleneck in manual services, multiple buyers may send consultation messages at the same time, which often makes it difficult for manual services to reply in a timely manner, thus Cause a decline in user experience and even lose users. [0003] Therefore, in order to solve the problem of high success and untimely service of manual customer service, chat systems s...

AI Technical Summary

Problems solved by technology

[0006] The intention recognition system of the existing technology needs a large amount of training data to obtain a robust model or system. Data is almost impossible, which makes it difficult to accurately identify intent

[0007] In the current popular bert model, self-supervised learning is required on the basis of a large amount of corpus. If a small amount of sample learning is lacking in data, the accuracy rate will also drop, and its generalization ability is relatively weak.

Images