Neural Network Text Classification Trend Regression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Text classification using neural networks faces challenges in accurately classifying data with trend relationships between categories, such as 'excellent', 'good', and 'poor', and often prioritizes overall accuracy over the correctness of individual data points, leading to low classification accuracy for critical data.
Innovation Solution
A method for training classification neural networks that involves acquiring a regression result representing the classification trend of sample text data, inputting this result along with the sample text data into a second target neural network, and adjusting parameters based on the difference between predicted and true classification results to achieve convergence, thereby improving classification accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional neural network classification is used, then overall classification accuracy can be ensured, but classification accuracy of critical data with trend relationships deteriorates
Solution Approach 1:
The patent segments the classification problem into two parts: a regression network that predicts trend scores for data points, and a classification network that uses these scores as additional features. This segmentation allows the system to handle trend relationships separately while maintaining overall classification performance, thereby improving critical data accuracy without excessive complexity increase.
Solution Approach 2:
The patent introduces regression results as an intermediary element between the input data and the classification network. These regression results serve as mediators that capture trend relationships and provide additional contextual information to the classification network, enabling more accurate classification of critical data points with trend characteristics.
2Measurement precision
If regression results are added as new dimensions to sample text vectors, then classification accuracy improves, but computational complexity increases
Solution Approach 1:
The patent performs regression analysis as a preliminary action before the main classification task. By pre-computing trend scores and integrating them into the feature vectors beforehand, the system prepares enriched input data that improves classification accuracy while allowing the main classification network to operate efficiently on pre-processed features.
3Measurement precision
If all data is treated equally in training, then overall accuracy can be maintained, but accuracy of individual critical data points deteriorates
Solution Approach 1:
The patent applies local quality by using regression results to provide different weighted contributions to the classification loss for different data points. Critical data points with significant trend relationships receive more attention through their regression-derived features, while other data points are processed normally, thereby improving individual critical point accuracy without requiring complete reworking of the training process.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Provided are a method and apparatuses for training a classification neural network, a text classification method and apparatus and an electronic device. The method includes: acquiring a regression result of sample text data, which is determined based on a pre-constructed first target neural network and represents a classification trend of the sample text data; inputting the sample text data and the regression result to a second target neural network; obtaining a predicted classification result of each piece of sample text data based on the second target neural network; adjusting a parameter of the second target neural network according to a difference between the predicted classification result and a true value of a corresponding category; and obtaining a trained second target neural network after a change of network loss meets a convergence condition. The second target neural network is trained better, and accuracy of subsequent text data classification is improved.