Method and system for training text classification model for spoken language interaction

Abstract

The embodiment of the present invention provides a training method for a text classification model for oral interaction. The method includes: acquiring a training set of spoken text corpus and dialogue history context information; expanding the spoken text corpus training set through the dialogue history context information to enrich the training set of spoken text corpus; establishing a text classification model based on a bidirectional long-term and short-term memory network , through the dialogue history context information and the expanded oral text corpus training set, the text classification model is trained, so that the text classification model can learn the field classification of the spoken text through the dialogue history context information. The embodiment of the present invention also provides a training system for a text classification model for spoken language interaction. The embodiment of the present invention determines the dialogue history context information, constructs a large number of virtual dialogue texts, and makes up for the lack of corpus; the dialogue history context information is used as a part of the input of the training model, and the dialogue history context information helps the model to improve the accuracy of domain classification.

Description

technical field [0001] The invention relates to the field of intelligent speech dialogue, in particular to a training method and system for a text classification model for spoken language interaction. Background technique [0002] In the text classification of spoken language interaction, a large number of artificially annotated corpora are usually used to train a deep learning model, and the model can automatically obtain text features. After the model outputs the results, it is also necessary to combine the last round of dialogue state design rules to select the final domain output. [0003] In the process of realizing the present invention, the inventor found that there are at least the following problems in the related art: [0004] The text classification method based on feature engineering requires labor-intensive design of text features. The quality of feature design restricts the final performance of the model, and the features used in this method often have problems...

AI Technical Summary

Problems solved by technology

[0006] In order to at least solve the time-consuming and labor-intensive task of manually designing text features in the prior art; after the classification model obtains the results, it is necessary to manually design rules to judge the final domain, which is not only time-consuming and labor-intensive, but also not flexible enough; dialog information can help the model judge the domain, Improve the accuracy of domain classification, but there is no problem of adding dialogue information in the existing method model

Images