Personalized Jargon Identification for Enhanced Interdisciplinary Communication

TL;DR

This paper explores personalized methods for identifying scientific jargon by accounting for individual researchers' backgrounds, using a new dataset and comparing supervised and prompt-based approaches to improve interdisciplinary communication.

Contribution

It introduces a dataset of individual familiarity annotations and demonstrates that personalized, prompt-based models outperform traditional corpus-level methods in jargon identification.

Findings

Prompt-based methods with personal publications achieve highest accuracy.

Zero-shot prompting provides a strong baseline.

Personalized features significantly improve jargon detection.

Abstract

Scientific jargon can impede researchers when they read materials from other domains. Current methods of jargon identification mainly use corpus-level familiarity indicators (e.g., Simple Wikipedia represents plain language). However, researchers' familiarity of a term can vary greatly based on their own background. We collect a dataset of over 10K term familiarity annotations from 11 computer science researchers for terms drawn from 100 paper abstracts. Analysis of this data reveals that jargon familiarity and information needs vary widely across annotators, even within the same sub-domain (e.g., NLP). We investigate features representing individual, sub-domain, and domain knowledge to predict individual jargon familiarity. We compare supervised and prompt-based approaches, finding that prompt-based methods including personal publications yields the highest accuracy, though zero-shot prompting provides a strong baseline. This research offers insight into features and methods to integrate personal data into scientific jargon identification.