Exploring the Role of Semantic Parsing on Downstream Tasks for Large Language Models
Abstract: Semantic Parsing focuses on converting sentences into structured forms. While previous studies show its benefits for smaller models, the impact on Large Language Models (LLMs) remains under explored. Our paper explores whether integrating Semantic Parsing can enhance LLMs' performance in downstream tasks. Unlike prior approaches, we propose SENSE, adding semantic parsing hint instead results into prompt and find that this approach consistently improves performance across tasks, highlighting the potential of semantic information integration in enhancing LLM capabilities.
Paper Type: Short
Research Area: Semantics: Lexical and Sentence-Level
Research Area Keywords: Semantic Parsing, Large Language Models
Contribution Types: Model analysis & interpretability, NLP engineering experiment
Languages Studied: English, Chinese, Germany
Submission Number: 5565
Loading