# On the Granularity of Explanations in Model Agnostic NLP Interpretability

arXiv: https://arxiv.org/abs/2012.13189 · HTML: https://mlsanity.com/paper/2012.13189/

## Key facts

- Cited by 3 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 24 Dec 2020.

## Built on by (most cited first, 3 of 3)

- [An Attention Matrix for Every Decision: Faithfulness-based Arbitration Among Multiple Attention-Based Interpretations of Transformers in Text Classification](https://mlsanity.com/paper/2209.10876/) (2022)
- [The FIX Benchmark: Extracting Features Interpretable to eXperts](https://mlsanity.com/paper/2409.13684/) (2024)
- [Explaining Black-box Model Predictions via Two-level Nested Feature Attributions with Consistency Property](https://mlsanity.com/paper/2405.14522/) (2024)

## Abstract (arXiv)

Current methods for Black-Box NLP interpretability, like LIME or SHAP, are based on altering the text to interpret by removing words and modeling the Black-Box response. In this paper, we outline limitations of this approach when using complex BERT-based classifiers: The word-based sampling produces texts that are out-of-distribution for the classifier and further gives rise to a high-dimensional search space, which can't be sufficiently explored when time or computation power is limited. Both of these challenges can be addressed by using segments as elementary building blocks for NLP interpretability. As illustration, we show that the simple choice of sentences greatly improves on both of these challenges. As a consequence, the resulting explainer attains much better fidelity on a benchmark classification task.