Text feature extraction method and device, text feature extraction model optimization method and device, medium and apparatus

Abstract

The invention provides a text feature extraction method and device, a text feature extraction model optimization method and device, a medium and an apparatus. The optimization method of the text feature extraction model comprises the following steps: acquiring a first feature matrix corresponding to each of the angles of view of the text data with multi-angle of view characteristics; fusing firstfeature matrices corresponding to each of the perspectives according to network attribute information corresponding to each of the perspectives to obtain a second feature matrix of the text data; whenthe second feature matrix satisfies a predetermined condition, network attribute information corresponding to each of the angles of view optimized by the text feature extraction model is outputted, and the second feature matrix is outputted as a network data matrix optimized by the text feature extraction model. The invention improves the accuracy and precision of the application tasks such as text semantic analysis and classification.

Description

technical field [0001] Embodiments of the present invention relate to the field of information processing technology, and more specifically, embodiments of the present invention relate to text feature extraction and extraction model optimization methods, media, devices, and equipment. Background technique [0002] This section is intended to provide a background or context for implementations of the invention that are recited in the claims. The descriptions herein are not admitted to be prior art by inclusion in this section. [0003] At present, the technical methods of deep learning have been widely applied to various fields of image and text processing. Among them, the deep learning technology represented by autoencoder, convolutional neural network, recurrent neural network and long-term short-term memory network has very good feature learning ability. , is widely used in practical projects of image classification and text semantic detection in academia and industry. H...

AI Technical Summary

Problems solved by technology

However, this technical method does not use the encoding network of the autoencoder to eliminate the differences between different features, and the features learned by training also lack non-negative interpretability.

In addition, this method can only obtain the required data features by batch training in the form of transduction, and cannot directly use the model parameters to inductively obtain the features after data fusion when new data arrives.

Images