Investigating Reasons for Disagreement in Natural Language Inference

TL;DR

This paper explores the causes of disagreement in natural language inference annotations, proposing a taxonomy of sources and comparing modeling approaches to detect ambiguous items, ultimately improving interpretability.

Contribution

It introduces a detailed taxonomy of disagreement sources in NLI and evaluates two modeling methods, finding multilabel classification more effective for capturing ambiguity.

Findings

Multilabel classification outperforms 4-way classification in detecting ambiguous items.

Disagreements stem from sentence meaning uncertainty, annotator biases, and task artifacts.

A taxonomy of 10 disagreement sources was developed.

Abstract

We investigate how disagreement in natural language inference (NLI) annotation arises. We developed a taxonomy of disagreement sources with 10 categories spanning 3 high-level classes. We found that some disagreements are due to uncertainty in the sentence meaning, others to annotator biases and task artifacts, leading to different interpretations of the label distribution. We explore two modeling approaches for detecting items with potential disagreement: a 4-way classification with a "Complicated" label in addition to the three standard NLI labels, and a multilabel classification approach. We found that the multilabel classification is more expressive and gives better recall of the possible interpretations in the data.