MonaLog: a Lightweight System for Natural Language Inference Based on Monotonicity

TL;DR

MonaLog is a lightweight natural logic inference engine leveraging monotonicity calculus, which effectively enhances BERT's performance on NLI tasks by generating high-quality training data.

Contribution

The paper introduces MonaLog, a simple yet effective logic-based NLI system based on natural logic and monotonicity, and demonstrates its utility in data augmentation for BERT.

Findings

MonaLog performs competitively on the SICK benchmark.

Combining MonaLog with BERT improves NLI accuracy.

MonaLog can generate high-quality training data for large models.

Abstract

We present a new logic-based inference engine for natural language inference (NLI) called MonaLog, which is based on natural logic and the monotonicity calculus. In contrast to existing logic-based approaches, our system is intentionally designed to be as lightweight as possible, and operates using a small set of well-known (surface-level) monotonicity facts about quantifiers, lexical items and tokenlevel polarity information. Despite its simplicity, we find our approach to be competitive with other logic-based NLI models on the SICK benchmark. We also use MonaLog in combination with the current state-of-the-art model BERT in a variety of settings, including for compositional data augmentation. We show that MonaLog is capable of generating large amounts of high-quality training data for BERT, improving its accuracy on SICK.