nasa-smd-ibm-v0.1_NER_DEAL

INDUS - NER-DEAL

Indus-NER-DEAL (nasa-smd-ibm-v0.1_NER_DEAL) is a RoBERTa-based, Encoder-only transformer model, domain-adapted for NASA Science Mission Directorate (SMD) applications. It's fine-tuned on scientific journals and articles relevant to NASA SMD, aiming to enhance natural language technologies like information retrieval and intelligent search.
This specific fork was finetuned on SciX Digital Library (https://scixplorer.org/, formerly NASA-ADS) proprietary data to label text with DEAL labels (https://ui.adsabs.harvard.edu/WIESP/2022/LabelDefinitions)

Usage

from transformers import AutoModelForTokenClassification, AutoTokenizer
INDUS_NER_DEAL = AutoModelForTokenClassification.from_pretrained(pretrained_model_name_or_path='adsabs/nasa-smd-ibm-v0.1_NER_DEAL',
                                                                 revision=None,
                                                                )

INDUS_tokenizer = AutoTokenizer.from_pretrained(pretrained_model_name_or_path='adsabs/nasa-smd-ibm-v0.1_NER_DEAL',
                                                 do_lower_case=False,
                                                )

Model Details

• Base Model: RoBERTa
• Tokenizer: Custom
• Parameters: 125M

Training Data

• 5K acknowledgements and full-text fragments from astronomy papers provided by NASA-SciX with manually tagged astronomical facilities and other entities of interest (e.g., celestial objects).
• approximately 1.6M words

<!-- ## Note --> <!-- ## Citation --> <!-- If you find this work useful, please cite using the following bibtex citation: --> <!-- ## Disclaimer -->