A text classification method and device

Abstract

The invention discloses a text classification method and device. The method comprises the following steps: utilizing each unlabeled linguistic data training word vector model in a corpus to obtain a target word vector model; according to the target word vector model, carrying out word expansion on a preset keyword corresponding to an appointed classification category to obtain a phrase set corresponding to the appointed classification category; according to the corpus, independently training a classifier for each phrase in the phrase set to obtain a target classifier independently corresponding to each phrase; according to a preset validation set, carrying out classification accuracy verification on the target classifier corresponding to each phrase, and selecting the phrase of which the classification accuracy conforms to a first set condition as a target phrase; and according to the target phrase contained in each piece of linguistic data in the corpus, selecting the linguistic data which meets a second set condition, labelling the linguistic data as a positive example sample of the appointed classification category, and adding the positive example sample into a training set. By use of the scheme of the invention, when the training set is generated for the first time, only multiple keywords need to be selected, a plurality of verification samples need to be labelled, and therefore, the new linguistic data can be subsequently automatically labelled.

Description

technical field [0001] The present application relates to the technical field of sample classification, and more specifically, to a text classification method and device. Background technique [0002] Text classification means that the computer automatically classifies and marks the text collection according to a certain classification system. Text classification usually includes the following steps: establishment of a classification system, obtaining a labeled training set, training a classification model, and using the trained classification model to predict the category of new text. [0003] Among them, the process of obtaining the labeled training set is to perform category labeling on the corpus of unknown categories, and the labeled corpus forms the training set. The size of the training set and the accuracy of the corpus annotation will directly affect the recognition accuracy of the trained classification model. The way of obtaining labeled training sets in the pri...

AI Technical Summary

Problems solved by technology

[0005] In view of this, this application provides a text classification method and device for solving the problem of wasting a lot of human resources in the existing manual labeling training set

Images