Method for constructing text classification system

Abstract

Provided is a method of constructing a text classification system in which text is associated with topics, and the method comprises the steps: constructing coarse classification tags of the text classification system, wherein the coarse classification tags are formed from topics and each having a topic list; expanding a topic list of the coarse classification labels; and constructing a fine classification tag of the coarse classification tag based on the extended topic list, at least a portion of the fine classification tag being formed by a keyword of the text associated with the topic list.According to the method, the topic labels associated with the texts are used for automatically constructing a coarse-to-fine multi-granularity text classification system, so the application requirements in different scenes can be supported, and the coarse-to-fine hierarchical extension of the category system can be automatically carried out.

Description

Technical field [0001] The present disclosure relates to the field of computer technology, and more specifically, to a method, device, electronic device, and computer-readable storage medium for constructing a text classification system. Background technique [0002] Nowadays, there is a huge amount of text information on the Internet. How to rationally arrange and use this information at different granularities to achieve different goals is a challenging and significant task. Generally speaking, to achieve this goal is usually to use tags of different granularities to classify and integrate information, and to retrieve corresponding information according to requirements in combination with tags of corresponding granularities during use. [0003] Existing text classification technologies are based on different ways of extracting text information features, and are mainly divided into three methods based on expert rules, based on artificial statistical features, and based on deep lea...

AI Technical Summary

Problems solved by technology

However, it ignores the effect of word order on the representation of text information. The feature latitude is high and the sparsity is strong, which is not conducive to the learning of classifiers.

[0006] The difficulty of deep learning-based methods lies in text representation, which is limited to one-dimensional label classification learning. For example, for a piece of text such as "What are the breaking news about the Huawei Mate 30 series, and what do you expect from the Mate 30 series?", most scenarios The problem to be solved is that it is correct to classify as "digital", or it is correct to classify as "mobile phone" under another classification system, but it lacks the ability to classify different granularity tags at the same time

Images