TERMinator: A system for scientific texts processing

TL;DR

This paper introduces TERMinator, a system for extracting entities and semantic relations from scientific texts, along with a new annotated dataset, highlighting the impact of language models and heuristics on extraction performance.

Contribution

The paper presents a novel system and dataset for scientific text processing, and compares the effectiveness of language models and heuristics in relation extraction.

Findings

Language models pre-trained on the target language do not always outperform others.

Heuristic approaches can enhance relation extraction quality.

The system and dataset are publicly available for research use.

Abstract

This paper is devoted to the extraction of entities and semantic relations between them from scientific texts, where we consider scientific terms as entities. In this paper, we present a dataset that includes annotations for two tasks and develop a system called TERMinator for the study of the influence of language models on term recognition and comparison of different approaches for relation extraction. Experiments show that language models pre-trained on the target language are not always show the best performance. Also adding some heuristic approaches may improve the overall quality of the particular task. The developed tool and the annotated corpus are publicly available at https://github.com/iis-research-team/terminator and may be useful for other researchers.