Exploring the Influence of Relevant Knowledge for Natural Language Generation Interpretability

TL;DR

This study demonstrates that relevant external knowledge significantly enhances the coherence and concept coverage of natural language generation, emphasizing the importance of interpretability and knowledge integration in NLG systems.

Contribution

The paper introduces KITGI, a new benchmark for evaluating knowledge influence in NLG, and provides a three-stage interpretability method to assess the role of external knowledge.

Findings

Full knowledge yields 91% correctness in generated sentences.

Removing key knowledge drops performance to 6%.

External knowledge is essential for coherent and comprehensive NLG.

Abstract

This paper explores the influence of external knowledge integration in Natural Language Generation (NLG), focusing on a commonsense generation task. We extend the CommonGen dataset by creating KITGI, a benchmark that pairs input concept sets with retrieved semantic relations from ConceptNet and includes manually annotated outputs. Using the T5-Large model, we compare sentence generation under two conditions: with full external knowledge and with filtered knowledge where highly relevant relations were deliberately removed. Our interpretability benchmark follows a three-stage method: (1) identifying and removing key knowledge, (2) regenerating sentences, and (3) manually assessing outputs for commonsense plausibility and concept coverage. Results show that sentences generated with full knowledge achieved 91\% correctness across both criteria, while filtering reduced performance drastically to 6\%. These findings demonstrate that relevant external knowledge is critical for maintaining both coherence and concept coverage in NLG. This work highlights the importance of designing interpretable, knowledge-enhanced NLG systems and calls for evaluation frameworks that capture the underlying reasoning beyond surface-level metrics.