An Interpretable Few-Shot Text Classification Model Based on Graph Neural Networks and Knowledge Graphs

few-shot data scarcity: Existing few-shot text classification methods face a data scarcity challenge that creates a data bottleneck in few-shot learning.

insufficient model interpretability: Existing few-shot text classification methods suffer from insufficient interpretability for classification decisions.

few-shot data scarcity: Existing few-shot text classification methods face a data scarcity challenge that creates a data bottleneck in few-shot learning.

insufficient model interpretability: Existing few-shot text classification methods suffer from insufficient interpretability for classification decisions.

ARExplainer: The paper proposes ARExplainer, an interpretable few-shot text classification method using data and reasoning augmentation.

LLM-based sample augmentation: The method uses LLM generalization capability to expand the diversity of training samples.

knowledge-graph reasoning engine: The method constructs a knowledge graph-driven reasoning engine that combines a Graph Attention Network to extract verifiable symbolic reasoning paths.

prompt-based explanation generation: The method uses a prompt-based explanation generator to produce concise and clear natural language explanations.

ARExplainer: The paper proposes ARExplainer, an interpretable few-shot text classification method using data and reasoning augmentation.

improved 1-shot classification performance:

more human-understandable explanations:

improved 1-shot classification performance: