AIObioEnts: All-in-one biomedical entities
Biomedical named-entity recognition following the all-in-one NER (AIONER) scheme introduced by Luo et al.. This is a straightforward Hugging-Face-compatible implementation without using a decoding head for ease of integration with other pipelines.
For full details, see the main GitHub repository
Anatomical biomedical entities
We have followed the original AIONER training pipeline based on the BioRED dataset along with additional BioRED-compatible datasets for set of core entities (Gene, Disease, Chemical, Species, Variant, Cell line), which we have fine-tuned using a modified version of the latest release of the AnatEM corpus, and a subset of entities that are of interest to us: cell, cell component, tissue, muti-tissue structure, and organ, along with the newly-introduced cancer. This model corresponds to the implementation based on BioLinkBERT-large
F1 scores
The F1 scores on the test set of this modified dataset are shown below:
| BioLink-large | |
|---|---|
| Cell | 89.28 |
| Cell component | 81.23 |
| Tissue | 74.49 |
| Cancer | 88.35 |
| Organ | 81.02 |
| Multi-tissue structure | 72.98 |
| Overall | 84.39 |
Usage
The model can be directly used from HuggingFace in a NER pipeline. However, we note that:
- The model was trained on sentence-level data, and it works best when the input is split
- Each sentence to tag must be surrounded by the flag corresponding to the entity type one wishes to identify, as in:
<entity_type>sentence</entity_type>. In the case of this fine-tuned model, the entity type should be'ALL'. - Since additional
'O'labels are used in the AIONER scheme, the outputs should be postprocessed before aggregating the tags
We provide helper functions to tag individual texts in the main repository
from tagging_fn import process_one_text
from transformers import pipeline
pipe = pipeline('ner', model='SIRIS-Lab/AIObioEnts-AnatEM-biolink-large', aggregation_strategy='none', device=0)
process_one_text(text_to_tag, pipeline=pipe, entity_type='ALL')