Understanding Dataset Difficulty with $\mathcal{V}$-Usable Information

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Ethayarajh, Kawin, Choi, Yejin, Swayamdipta, Swabha
Format: Preprint
Published: 2021
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917999959605248
author Ethayarajh, Kawin
Choi, Yejin
Swayamdipta, Swabha
author_facet Ethayarajh, Kawin
Choi, Yejin
Swayamdipta, Swabha
contents Estimating the difficulty of a dataset typically involves comparing state-of-the-art models to humans; the bigger the performance gap, the harder the dataset is said to be. However, this comparison provides little understanding of how difficult each instance in a given distribution is, or what attributes make the dataset difficult for a given model. To address these questions, we frame dataset difficulty -- w.r.t. a model $\mathcal{V}$ -- as the lack of $\mathcal{V}$-$\textit{usable information}$ (Xu et al., 2019), where a lower value indicates a more difficult dataset for $\mathcal{V}$. We further introduce $\textit{pointwise $\mathcal{V}$-information}$ (PVI) for measuring the difficulty of individual instances w.r.t. a given distribution. While standard evaluation metrics typically only compare different models for the same dataset, $\mathcal{V}$-$\textit{usable information}$ and PVI also permit the converse: for a given model $\mathcal{V}$, we can compare different datasets, as well as different instances/slices of the same dataset. Furthermore, our framework allows for the interpretability of different input attributes via transformations of the input, which we use to discover annotation artefacts in widely-used NLP benchmarks.
format Preprint
id arxiv_https___arxiv_org_abs_2110_08420
institution arXiv
publishDate 2021
record_format arxiv
spellingShingle Understanding Dataset Difficulty with $\mathcal{V}$-Usable Information
Ethayarajh, Kawin
Choi, Yejin
Swayamdipta, Swabha
Computation and Language
Artificial Intelligence
Machine Learning
Estimating the difficulty of a dataset typically involves comparing state-of-the-art models to humans; the bigger the performance gap, the harder the dataset is said to be. However, this comparison provides little understanding of how difficult each instance in a given distribution is, or what attributes make the dataset difficult for a given model. To address these questions, we frame dataset difficulty -- w.r.t. a model $\mathcal{V}$ -- as the lack of $\mathcal{V}$-$\textit{usable information}$ (Xu et al., 2019), where a lower value indicates a more difficult dataset for $\mathcal{V}$. We further introduce $\textit{pointwise $\mathcal{V}$-information}$ (PVI) for measuring the difficulty of individual instances w.r.t. a given distribution. While standard evaluation metrics typically only compare different models for the same dataset, $\mathcal{V}$-$\textit{usable information}$ and PVI also permit the converse: for a given model $\mathcal{V}$, we can compare different datasets, as well as different instances/slices of the same dataset. Furthermore, our framework allows for the interpretability of different input attributes via transformations of the input, which we use to discover annotation artefacts in widely-used NLP benchmarks.
title Understanding Dataset Difficulty with $\mathcal{V}$-Usable Information
topic Computation and Language
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2110.08420