Abstract:With continuous development of artificial intelligence, text classification has gradually changed from a knowledge-based method to a method based on statistics and machine learning. Among them, it is a very important and efficient way to classify text based on the convolutional neural network (CNN) model. Text data are a kind of sequence data, while time sequentiality of the general text data is relatively weak, so text classification is usually less relevant to the sequential structure of the full text. Therefore, CNN-based text classification has gradually become a research hotspot when dealing with issues of text classification. For machine learning, especially deep learning, model interpretability has increasingly become the focus of academic research and industrial applications, and also become a key issue for further development and application of deep learning technology. Therefore, we recommend using the backtracking analysis method to conduct in-depth research on deep learning models. This paper proposes an analysis method for interpretability of a CNN text classification model. The method proposed by us can perform multi-angle analysis on the discriminant results of multi-classified text and multi-label classification tasks through backtracking analysis on model prediction results. Finally, the analysis results of the model can be displayed using visualization technology from multiple dimensions based on interpretability. The representative data set IMDB (Internet Movie Database) in text classification is verified by examples, and the results show that the model can be effectively analyzed when using our method.

On the Interpretation of Convolutional Neural Networks for Text Classification

A Pixel-Level Explainable Approach of Convolutional Neural Networks and Its Application

Understanding Convolutional Neural Networks for Text Classification

Interpretable Text Classification Using CNN and Max-pooling

An Analysis Method for Interpretability of CNN Text Classification Model

Deeper Interpretability of Deep Networks

Hybrid CNN -Interpreter: Interpret local and global contexts for CNN-based Models

Interpretable CNNs for Object Classification

Text Classification Method Based On Convolution Neural Network

Exploring Frequency Domain Interpretation of Convolutional Neural Networks

Character-level Convolutional Networks for Text Classification

Interpreting the decisions of CNNs via influence functions

What Does a TextCNN Learn?

Interpretable Deep Convolutional Neural Networks via Meta-learning

Visualizing Surrogate Decision Trees of Convolutional Neural Networks

Feature-Based Interpretation of Image Classification With the Use of Convolutional Neural Networks

PICNN: A Pathway towards Interpretable Convolutional Neural Networks

FICNN: A Framework for the Interpretation of Deep Convolutional Neural Networks

A convolutional attention model for text classification

Interpreting Adversarially Trained Convolutional Neural Networks

Transformable Convolutional Neural Network for Text Classification