Fumeus: A family of Python tools for text mining with smoke terms
David M. Goldberg, Richard J. Gruss, Alan S. Abrahams · Software Impacts · 2022
Synthesizing meaningful insights from voluminous textual datasets is complex and challenging. The task is especially difficult for sparse target classes. Recent works have proposed "smoke terms," or machine-learned words and phrases prevalent in a target class. Smoke terms may be utilized to rank or sort text, or they may serve as features in follow-on machine learning models. This paper introduces Fumeus, a family of Python-based smoke term analysis tools. We provide functionality to generate new smoke terms and to use existing smoke term dictionaries to rank or sort datasets. These analyses have numerous academic, regulatory, and industry applications.