Extracting Fine-Grained Knowledge Graphs of Scientific Claims: Dataset and Transformer-Based Results
This work addresses the need for more detailed scientific information extraction, particularly for researchers in social and behavioral sciences and biomedicine, by providing a dataset with higher label density and fine-grained attributes, though it is incremental in building upon existing transformer-based approaches.
The authors tackled the problem of extracting fine-grained knowledge graphs from scientific claims by introducing SciClaim, a dataset with 12,738 labels that captures detailed associations and attributes, and they demonstrated promising results using transformer-based methods for joint entity and relation extraction.
Recent transformer-based approaches demonstrate promising results on relational scientific information extraction. Existing datasets focus on high-level description of how research is carried out. Instead we focus on the subtleties of how experimental associations are presented by building SciClaim, a dataset of scientific claims drawn from Social and Behavior Science (SBS), PubMed, and CORD-19 papers. Our novel graph annotation schema incorporates not only coarse-grained entity spans as nodes and relations as edges between them, but also fine-grained attributes that modify entities and their relations, for a total of 12,738 labels in the corpus. By including more label types and more than twice the label density of previous datasets, SciClaim captures causal, comparative, predictive, statistical, and proportional associations over experimental variables along with their qualifications, subtypes, and evidence. We extend work in transformer-based joint entity and relation extraction to effectively infer our schema, showing the promise of fine-grained knowledge graphs in scientific claims and beyond.