Explaining Predictions of Non-Linear Classifiers in NLP

@inproceedings{Arras2016ExplainingPO,
  title={Explaining Predictions of Non-Linear Classifiers in NLP},
  author={Leila Arras and F. Horn and Gr{\'e}goire Montavon and K. M{\"u}ller and W. Samek},
  booktitle={Rep4NLP@ACL},
  year={2016}
}
  • Leila Arras, F. Horn, +2 authors W. Samek
  • Published in Rep4NLP@ACL 2016
  • Computer Science, Mathematics
  • Layer-wise relevance propagation (LRP) is a recently proposed technique for explaining predictions of complex non-linear classifiers in terms of input variables. In this paper, we apply LRP for the first time to natural language processing (NLP). More precisely, we use it to explain the predictions of a convolutional neural network (CNN) trained on a topic categorization task. Our analysis highlights which words are relevant for a specific prediction of the CNN. We compare our technique to… CONTINUE READING
    62 Citations
    Looking Deeper into Deep Learning Model: Attribution-based Explanations of TextCNN
    • 2
    • Highly Influenced
    • PDF
    Evaluating neural network explanation methods using hybrid documents and morphological prediction
    • 5
    • Highly Influenced
    • PDF
    Explaining Recurrent Neural Network Predictions in Sentiment Analysis
    • 149
    • PDF
    Evaluating Recurrent Neural Network Explanations
    • 23
    • PDF
    Evaluating neural network explanation methods using hybrid documents and morphological agreement
    • 5
    • Highly Influenced
    Human-grounded Evaluations of Explanation Methods for Text Classification
    • 9
    • PDF
    "What is relevant in a text document?": An interpretable machine learning approach
    • 131
    • PDF
    Evaluating neural network explanation methods using hybrid documents and morphosyntactic agreement
    • 35
    • Highly Influenced
    • PDF

    References

    SHOWING 1-10 OF 25 REFERENCES
    Visualizing and Understanding Neural Models in NLP
    • 389
    • PDF
    Convolutional Neural Networks for Sentence Classification
    • 7,121
    • PDF
    On Pixel-Wise Explanations for Non-Linear Classifier Decisions by Layer-Wise Relevance Propagation
    • 1,166
    • PDF
    Analyzing Classifiers: Fisher Vectors and Deep Neural Networks
    • 109
    • PDF
    Explaining nonlinear classification decisions with deep Taylor decomposition
    • 500
    • PDF
    Visualizing and Understanding Convolutional Networks
    • 8,595
    • PDF
    Natural Language Processing (Almost) from Scratch
    • 5,637
    • PDF
    Interpreting individual classifications of hierarchical networks
    • 54
    • PDF
    Evaluating the Visualization of What a Deep Neural Network Has Learned
    • 404
    • PDF
    Understanding Representations Learned in Deep Architectures
    • 85
    • PDF