Intention corpus generation method and device based on machine learning, and readable storage medium

Abstract

The invention relates to the technical field of artificial intelligence, and provides an intention corpus generation method and device based on machine learning, electronic equipment and a computer readable storage medium, and the method comprises the steps: marking a statement corresponding to each intention as an intention seed corpus; labeling the intention seed corpora as training intention corpora through a pre-trained labeling model, and obtaining a group of intention seed corpora with the highest similarity through a preset similarity formula; performing labeling processing on the intention seed corpus with the highest similarity through a labeling model to obtain a training intention corpus; training the constructed intention corpus generation model through the training intention corpus until the intention corpus generation model is converged to a preset range; and processing the client expression statements of the man-machine conversation into different intention corpora through the trained intention corpus generation model. The invention mainly aims to automatically generate more intention corpora with different expressions through an intention corpus generation model for a part of labeled intention corpora.

Description

technical field [0001] The present invention relates to the technical field of artificial intelligence, and in particular to a method, device, electronic device and computer-readable storage medium for generating intent corpus based on machine learning. Background technique [0002] At present, the dialogue robot uses the intention recognition model to recognize the intention of the customer's speech content. When the intention recognition model recognizes, it needs to learn a number of marked corpus; generally speaking, the customer's speech content has multiple intentions, and each intention corresponds to multiple The corpus (sentence) that expresses the intent; there are many ways to express an intent, and it is difficult for annotators to enumerate all the corpus. [0003] The traditional intention corpus generation method: first manually specify the keywords, and then generate a corpus by filling in the corpus template, or replace some keywords in the existing corpus t...

AI Technical Summary

Problems solved by technology

[0002] At present, the dialogue robot uses the intention recognition model to recognize the intention of the customer's speech content. When the intention recognition model recognizes, it needs to learn a number of marked corpus; generally speaking, the customer's speech content has multiple intentions, and each intention corresponds to multiple The corpus (sentence) that expresses the intention; there are many ways to express an intention, and it is difficult for annotators to enumerate all the corpus

[0003] The traditional intention corpus generation method: first manually specify the keywords, and then generate a corpus by filling in the corpus template, or replace some keywords in the existing corpus to form a new corpus. The corpus generated by this method only adds keywords and Combining information of different expressions does not increase the richness of the semantic expression itself, that is to say, no more semantic information is added, and manual sorting of keywords increases labor costs

[0004] Some use machine translation to translate the corpus into other languages ​​first, and then translate the translated text back to the original language of the corpus. Although this method does not require additional manual annotation, but because of the machine translation used, after two rounds of translation, the back-translation The corpus often has problems such as incoherent sentences and changes in semantic information.

Images