laurabernardy/RueBERTa
Updated • 2
Dataset consisting of 66.438 html-cleaned samples taken from the Dataset 9 of the Epstein Files.
Samples WITH Location Entity: 22.146 Samples WITHOUT Location Entity: 44.292
The dataset was labeled using a weak labeling approach using qwen3 32b.
It is BIO-tagged, B-ADDR and I-ADDR mark location entities. Normalized and manually corrected Labels are added as well.
Disclaimer: This work uses a subset of the publicly released Epstein files. The source material contains sensitive communications involving real individuals.