Short text classification method and system based on multi-feature fusion

Abstract

The invention relates to a short text classification method and system based on multi-feature fusion, and the method comprises the steps: firstly carrying out the preprocessing of a text, including word segmentation, stop word processing and feature selection; secondly, extracting features of the processed text through a word frequency and inverse word order method, a convolutional neural networkand a long-short-term memory network algorithm, and forming three feature vectors; then, fusing the three types of features, using an attention mechanism to weight the fused features, and highlightingimportant features; and finally, enabling the fusion features to pass through a classifier to obtain a short text classification result. A feature dictionary is established by using word frequency inverse word order features, and vectorization representation is performed on a text; features are extracted in combination with a filter and a long-short-term memory network, the three kinds of features are fused to enrich short text features, an attention mechanism is used for distributing weights, and the classification effect is stabilized.

Description

technical field [0001] The invention belongs to the field of natural language processing, and in particular relates to a short text classification method and system based on multi-feature fusion. Background technique [0002] With the advent of the era of big data, paper documents are rapidly changing to electronic and digital, and text classification has become one of the most common tasks in natural language processing. With the advancement of network technology and the development of electronic social media, a new type of text—short text has become an important form of network information. Short texts have become an important form for individuals to express their opinions and share information on online platforms. Short text data has a wide range of applications, such as questions raised by users in question answering systems, chat records in social network exchange forums, sentiment analysis on review sites, etc. Short text refers to short text, which is relative to do...

AI Technical Summary

Problems solved by technology

(4) The text format is not standard

However, the current short text classification technology mainly has the following problems: (1) Faced with specific scenarios and resources based on machine learning methods, the feature engineering steps depend on domain knowledge, the calculation efficiency is low, and it is difficult to extend to other scenarios

Moreover, using the bag-of-words feature to represent the text will lose the word order information of the text, and will cause the text feature dimension to be too high and the information content to be sparse, which only represents the very shallow content of the text.

Especially in the face of short text classification, the sparsity of features is serious, and it is difficult to have better results

(2) The current model cannot effectively solve the feature sparsity problem caused by short texts, and it is difficult to achieve better classification results

It cannot effectively solve the feature sparsity problem of short texts, and it is difficult to achieve better results in classification

Images