Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces TWIX, a two-stage end-to-end information extraction pipeline designed to tackle Named Entity Recognition (NER), Named Entity Recognition and Disambiguation (NERD), and Relation Extraction (RE) within the gut-brain axis domain. By leveraging a modular architecture, TWIX significantly enhances performance metrics, achieving superior precision and recall compared to baseline models and outperforming all other submissions in the GutBrainIE benchmark. The results underscore the effectiveness of the two-stage approach in addressing complex extraction tasks in a rapidly growing field of scientific literature.
TWIX outperforms existing benchmarks in named entity recognition and relation extraction, setting a new standard for information extraction in the gut-brain axis domain.
The exponential growth of scientific publications calls for automatic Information Extraction (IE) systems to support knowledge discovery. In this context, the GutBrainIE benchmark evaluates Named Entity Recognition (NER), Named Entity Recognition and Disambiguation (NERD), and Relation Extraction (RE) systems in the gut-brain axis domain. We propose Two-stage Workflow for Information eXtraction (TWIX), an end-to-end IE pipeline featuring three interconnected modules, each leveraging a two-stage framework to solve all four GutBrainIE subtasks. Evaluation on the development and test sets shows that our method substantially outperforms the baseline by a wide margin, while also ranking first among all participant submissions across all subtasks. These results indicate that the proposed two-stage pipeline effectively improves both precision and recall in practical settings.