Learning to Substitute Words with Model-based Score Ranking

2025-02-09

AI Summary

This work addresses the poor generalizability of intelligent lexical substitution methods caused by reliance on subjective human annotations. We propose the first unsupervised lexical substitution learning framework, abandoning manual annotation in favor of model-based quality scores—such as BARTScore—as surrogate supervision signals. Our approach constructs a word replacement distribution and statistical significance criterion grounded in these scores, and introduces a quality-score alignment loss for end-to-end self-supervised optimization. The key innovation lies in unifying model-based quality evaluation, statistical significance testing, and sequence-level substitution modeling into a single unsupervised training objective. Experiments demonstrate that our method significantly outperforms strong baselines—including BERT, BART, GPT-4, and LLaMA—across multiple benchmarks, substantially improving post-substitution sentence quality. The source code is publicly available.

Technology Category

• Machine Learning: Unsupervised & Self-Supervised Learning
• Natural Language Processing: Sentence-level Semantics, Textual Inference, etc.
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Smart word substitution aims to enhance sentence quality by improving word choices; however current benchmarks rely on human-labeled data. Since word choices are inherently subjective, ground-truth word substitutions generated by a small group of annotators are often incomplete and likely not generalizable. To circumvent this issue, we instead employ a model-based score (BARTScore) to quantify sentence quality, thus forgoing the need for human annotations. Specifically, we use this score to define a distribution for each word substitution, allowing one to test whether a substitution is statistically superior relative to others. In addition, we propose a loss function that directly optimizes the alignment between model predictions and sentence scores, while also enhancing the overall quality score of a substitution. Crucially, model learning no longer requires human labels, thus avoiding the cost of annotation while maintaining the quality of the text modified with substitutions. Experimental results show that the proposed approach outperforms both masked language models (BERT, BART) and large language models (GPT-4, LLaMA). The source code is available at https://github.com/Hyfred/Substitute-Words-with-Ranking.

Problem

• Enhances sentence quality via smart word substitution
• Eliminates reliance on human-labeled data for substitutions
• Optimizes model predictions to improve substitution quality

Innovation

• Model-based score ranking
• No human annotations required
• Optimized loss function alignment