Short text feature extraction method based on multi-feature factor fusion

Abstract

The invention relates to a short text feature extraction method based on multi-feature factor fusion, which comprises the following steps of: carrying out word segmentation and stop word removal processing on short text comments through a conjunctive word segmentation tool so as to construct a preliminary text feature word vector matrix; combining a traditional TF-IDF algorithm to carry out weightcalculation on the constructed feature word vector matrix by using an IDF obtain a weight vector matrix; introducing a feature word position influence factor and a part-of-speech feature factor, carrying out part-of-speech tagging on the preliminary text feature words one by one, and calculating the sum value of each feature word; multiplying the obtained sum values with a weight value corresponding to the conventional TF-IDF algorithm, to finally obtain a weight vector matrix of the optimized TF-IDF algorithm. According to the technical scheme provided by the invention, a word weight imbalance problem of the traditional TF-IDF algorithm can be solved to a certain extent, so that the text characteristic extraction accuracy is improved, and effective help is provided for emotion classification tasks.

Description

technical field [0001] The invention relates to the technical field of text mining, in particular to a short text feature extraction method based on the fusion of multiple feature factors. Background technique [0002] With the advancement of the Web3.0 era, Internet information has been increasingly integrated into people's lives. A large number of users express their opinions on an event or product on the Internet, and these comment information will greatly affect people's thinking and behavior under the time effect. At the same time, these comment information includes people's various emotional attitudes and emotional information, such as happiness, anger, sadness, joy, sadness or positive, neutral, and negative. Based on these comment information, other users can learn about group users' comments and opinions on an event or product through the network platform, so this information has huge potential mining value. In addition, during the rapid development of the Interne...

AI Technical Summary

Problems solved by technology

At the same time, this method does not take into account the front and rear positions of words and their own part-of-speech features, and there are certain important trade-offs between words that appear in the front position and words that appear in the back position, so that some errors are inevitable in the process of feature extraction

Images