SPOR: A Comprehensive and Practical Evaluation Method for Compositional Generalization in Data-to-Text Generation

TL;DR

This paper introduces SPOR, a comprehensive evaluation framework for assessing various aspects of compositional generalization in data-to-text generation, highlighting current models' deficiencies.

Contribution

It proposes a new evaluation method covering multiple manifestations of compositionality, applicable without manual annotations, and demonstrates its effectiveness on different datasets and models.

Findings

Models show deficiencies in multiple compositional aspects.

SPOR enables high-quality, annotation-free evaluation.

The framework highlights the need for further research in compositional generalization.

Abstract

Compositional generalization is an important ability of language models and has many different manifestations. For data-to-text generation, previous research on this ability is limited to a single manifestation called Systematicity and lacks consideration of large language models (LLMs), which cannot fully cover practical application scenarios. In this work, we propose SPOR, a comprehensive and practical evaluation method for compositional generalization in data-to-text generation. SPOR includes four aspects of manifestations (Systematicity, Productivity, Order invariance, and Rule learnability) and allows high-quality evaluation without additional manual annotations based on existing datasets. We demonstrate SPOR on two different datasets and evaluate some existing language models including LLMs. We find that the models are deficient in various aspects of the evaluation and need further improvement. Our work shows the necessity for comprehensive research on different manifestations of compositional generalization in data-to-text generation and provides a framework for evaluation.