Multi-Granular Text Encoding for Self-Explaining Categorization
It addresses the problem of providing intuitive evidence for predictions in text categorization, specifically for medical disease classification, with incremental improvements in method and efficiency.
The paper tackles self-explaining text categorization by using multi-granular ngrams organized hierarchically with a tree-structured LSTM, resulting in improved accuracy, efficiency, and compactness over BiLSTM and CNN baselines in medical disease classification.
Self-explaining text categorization requires a classifier to make a prediction along with supporting evidence. A popular type of evidence is sub-sequences extracted from the input text which are sufficient for the classifier to make the prediction. In this work, we define multi-granular ngrams as basic units for explanation, and organize all ngrams into a hierarchical structure, so that shorter ngrams can be reused while computing longer ngrams. We leverage a tree-structured LSTM to learn a context-independent representation for each unit via parameter sharing. Experiments on medical disease classification show that our model is more accurate, efficient and compact than BiLSTM and CNN baselines. More importantly, our model can extract intuitive multi-granular evidence to support its predictions.