Can I read Evaluating Feature Attribution in NLP on EtoBox?
Evaluating Feature Attribution in NLP by rifkyadi67 is a document available to read on EtoBox.
What is Evaluating Feature Attribution in NLP about?
The document presents TextFocus, a novel metric for evaluating the faithfulness of Feature Attribution methods in Natural Language Processing (NLP). It introduces the concept of textual mosaics, which serve as implicit ground truth for attribution, allowing for systematic comparison of six feature attribution methods across three classification tasks. The study finds that Integrated Gradients performs best in terms of faithfulness and computational efficiency, addressing a gap in the evaluation of XAI metho
- Author
- rifkyadi67
- Language
- EN