You can deploy this model and start earning money today!

Glasgow-AI4BioMed/
bioner_medmentions_st21pv_finegrain

Name: bioner_medmentions_st21pv_finegrain
Availability: InStock

microsoft/BiomedNLP-BiomedBERT-base-uncased-abstract-fulltext

finetune

mit

Views

No views yet

bioner_medmentions_st21pv_finegrain

This is a named entity recognition model fine-tuned from the microsoft/BiomedNLP-BiomedBERT-base-uncased-abstract-fulltext model. It predicts spans with 91 possible labels. The labels are Acquired Abnormality, Amino Acid Sequence, Amino Acid, Peptide, or Protein, Amphibian, Anatomical Abnormality, Anatomical Structure, Animal, Antibiotic, Bacterium, Biologic Function, Biologically Active Substance, Biomedical Occupation or Discipline, Biomedical or Dental Material, Bird, Body Location or Region, Body Part, Organ, or Organ Component, Body Space or Junction, Body Substance, Body System, Cell, Cell Component, Cell Function, Cell or Molecular Dysfunction, Chemical, Chemical Viewed Functionally, Chemical Viewed Structurally, Classification, Clinical Attribute, Congenital Abnormality, Diagnostic Procedure, Disease or Syndrome, Drug Delivery Device, Element, Ion, or Isotope, Embryonic Structure, Enzyme, Eukaryote, Experimental Model of Disease, Finding, Fish, Food, Fully Formed Anatomical Structure, Fungus, Gene or Genome, Genetic Function, Geographic Area, Hazardous or Poisonous Substance, Health Care Activity, Health Care Related Organization, Hormone, Human, Idea or Concept, Immunologic Factor, Indicator, Reagent, or Diagnostic Aid, Injury or Poisoning, Inorganic Chemical, Intellectual Product, Laboratory Procedure, Laboratory or Test Result, Mammal, Medical Device, Mental Process, Mental or Behavioral Dysfunction, Molecular Biology Research Technique, Molecular Function, Molecular Sequence, Neoplastic Process, Nucleic Acid, Nucleoside, or Nucleotide, Nucleotide Sequence, Organ or Tissue Function, Organic Chemical, Organism Function, Organization, Pathologic Function, Pharmacologic Substance, Physiologic Function, Plant, Population Group, Professional Society, Professional or Occupational Group, Receptor, Regulation or Law, Reptile, Research Activity, Self-help or Relief Organization, Sign or Symptom, Spatial Concept, Therapeutic or Preventive Procedure, Tissue, Vertebrate, Virus and Vitamin.

The code used for training this model can be found at https://github.com/Glasgow-AI4BioMed/bioner along with links to other biomedical NER models trained on well-known biomedical corpora. The source dataset information is below.

Example Usage

The code below will load up the model and apply it to the provided text. It uses a simple aggregation strategy to post-process the individual tokens into larger multi-token entities where needed.

python

from transformers import pipeline

# Load the model as part of an NER pipeline
ner_pipeline = pipeline("token-classification", 
                        model="Glasgow-AI4BioMed/bioner_medmentions_st21pv_finegrain",
                        aggregation_strategy="max")

# Apply it to some text
ner_pipeline("EGFR T790M mutations have been known to affect treatment outcomes for NSCLC patients receiving erlotinib.")

# Output:
# [ {"entity_group": "Gene or Genome", "score": 0.96229, "word": "egfr", "start": 0, "end": 4},
#   {"entity_group": "Genetic Function", "score": 0.91988, "word": "t790m mutations", "start": 5, "end": 20},
#   {"entity_group": "Neoplastic Process", "score": 0.99883, "word": "nsclc", "start": 51, "end": 56},
#   {"entity_group": "Pharmacologic Substance", "score": 0.99931, "word": "erlotinib", "start": 76, "end": 85} ]

Dataset Info

Source: The ST21pv version of MedMentions was downloaded from: https://github.com/chanzuckerberg/MedMentions/tree/master/st21pv

The dataset should be cited with: Mohan, Sunil, and Donghui Li. "MedMentions: A Large Biomedical Corpus Annotated with UMLS Concepts." Automated Knowledge Base Construction (AKBC), 2019, https://openreview.net/forum?id=SylxCx5pTQ. DOI: 10.24432/C5G59C

An overview of semantic types can be found at: https://www.nlm.nih.gov/research/umls/META3_current_semantic_types.html

Preprocessing: The training, validation and test splits were maintained from the original dataset. Concept identifiers (CUIs) were used to map each annotation to its associated UMLS entry to recover semantic types (from the MRSTY.RRF UMLS file). Semantic types provided in MedMentions were not used. Annotations were mapped to specific semantic types names using the Semantic Groups file available at: https://www.nlm.nih.gov/research/umls/knowledge_sources/semantic_network/index.html. This contrasts with the finegrained version that mapped annotations to semantic groups. The preprocessing script for this dataset is prepare_medmentions.py with the --finegrain flag.

Performance

The span-level performance on the test split for the different labels are shown in the tables below. The full performance results are available in the model repo in Markdown format for viewing and JSON format for easier loading. These include the performance at token level (with individual B- and I- labels as the token classifier uses IOB2 token labelling).

Label	Precision	Recall	F1-score	Support
Acquired Abnormality	0.273	0.240	0.255	50
Amino Acid Sequence	0.303	0.357	0.328	84
Amino Acid, Peptide, or Protein	0.258	0.277	0.267	166
Anatomical Abnormality	0.190	0.145	0.164	76
Anatomical Structure	0.160	0.308	0.211	13
Animal	0.605	0.742	0.667	93
Antibiotic	0.835	0.784	0.808	148
Bacterium	0.752	0.732	0.742	448
Biologic Function	0.341	0.369	0.355	157
Biologically Active Substance	0.586	0.640	0.612	2080
Biomedical Occupation or Discipline	0.437	0.444	0.441	196
Biomedical or Dental Material	0.369	0.452	0.406	197
Bird	0.810	0.819	0.814	83
Body Location or Region	0.386	0.461	0.420	232
Body Part, Organ, or Organ Component	0.565	0.604	0.584	1092
Body Space or Junction	0.272	0.311	0.290	90
Body Substance	0.542	0.731	0.622	212
Body System	0.554	0.511	0.532	90
Cell	0.680	0.737	0.708	924
Cell Component	0.589	0.637	0.612	311
Cell Function	0.484	0.599	0.535	499
Cell or Molecular Dysfunction	0.607	0.657	0.631	99
Chemical	0.348	0.333	0.340	72
Chemical Viewed Functionally	0.286	0.432	0.344	37
Chemical Viewed Structurally	0.443	0.427	0.435	82
Classification	0.520	0.544	0.532	309
Clinical Attribute	0.596	0.625	0.610	323
Congenital Abnormality	0.438	0.443	0.440	79
Diagnostic Procedure	0.672	0.648	0.660	735
Disease or Syndrome	0.757	0.774	0.766	2199
Element, Ion, or Isotope	0.713	0.657	0.684	385
Embryonic Structure	0.587	0.509	0.545	53
Enzyme	0.766	0.761	0.763	681
Eukaryote	0.745	0.793	0.768	397
Experimental Model of Disease	0.286	0.356	0.317	45
Finding	0.391	0.388	0.389	2759
Fish	1.000	0.947	0.973	19
Food	0.556	0.455	0.501	336
Fully Formed Anatomical Structure	0.000	0.000	0.000	1
Fungus	0.819	0.798	0.809	119
Gene or Genome	0.551	0.539	0.545	912
Genetic Function	0.598	0.646	0.621	652
Geographic Area	0.673	0.712	0.692	598
Hazardous or Poisonous Substance	0.513	0.522	0.518	293
Health Care Activity	0.487	0.458	0.472	1061
Health Care Related Organization	0.531	0.642	0.581	296
Hormone	0.806	0.746	0.775	189
Human	0.799	0.880	0.837	158
Idea or Concept	0.000	0.000	0.000	1
Immunologic Factor	0.674	0.606	0.638	434
Indicator, Reagent, or Diagnostic Aid	0.427	0.451	0.439	182
Injury or Poisoning	0.617	0.703	0.657	357
Inorganic Chemical	0.611	0.680	0.643	256
Intellectual Product	0.495	0.485	0.490	2075
Laboratory Procedure	0.445	0.452	0.448	908
Laboratory or Test Result	0.183	0.196	0.190	112
Mammal	0.778	0.838	0.807	456
Medical Device	0.434	0.437	0.435	355
Mental Process	0.546	0.546	0.546	740
Mental or Behavioral Dysfunction	0.710	0.774	0.741	518
Molecular Biology Research Technique	0.500	0.539	0.519	206
Molecular Function	0.504	0.555	0.528	719
Molecular Sequence	0.417	0.556	0.476	9
Neoplastic Process	0.761	0.745	0.753	918
Nucleic Acid, Nucleoside, or Nucleotide	0.331	0.450	0.381	109
Nucleotide Sequence	0.320	0.491	0.387	110
Organ or Tissue Function	0.482	0.425	0.452	247
Organic Chemical	0.396	0.464	0.427	511
Organism Function	0.462	0.518	0.488	471
Organization	0.270	0.442	0.335	77
Pathologic Function	0.541	0.541	0.541	669
Pharmacologic Substance	0.577	0.623	0.599	1258
Physiologic Function	0.283	0.286	0.284	182
Plant	0.603	0.618	0.610	403
Population Group	0.710	0.711	0.711	1263
Professional Society	0.000	0.000	0.000	7
Professional or Occupational Group	0.599	0.725	0.656	360
Receptor	0.614	0.686	0.648	271
Regulation or Law	0.182	0.125	0.148	16
Reptile	1.000	0.318	0.483	22
Research Activity	0.559	0.540	0.549	1653
Self-help or Relief Organization	0.000	0.000	0.000	2
Sign or Symptom	0.629	0.647	0.638	340
Spatial Concept	0.474	0.483	0.479	1282
Therapeutic or Preventive Procedure	0.609	0.617	0.613	2036
Tissue	0.562	0.525	0.543	259
Vertebrate	0.000	0.000	0.000	1
Virus	0.678	0.831	0.747	172
Vitamin	0.706	0.511	0.593	47
macro avg	0.507	0.525	0.512	40144
weighted avg	0.571	0.589	0.578	40144

Hyperparameters

Hyperparameter tuning was done with optuna and the hyperparameter_search functionality. 100 trials were run. Early stopping was applied during training. The best performing model was selected using the macro F1 performance on the validation set. The selected hyperparameters are in the table below.

Hyperparameter	Value
epochs	25.0
learning_rate	7.86794379743531e-05
per_device_train_batch_size	16
weight_decay	0.06816454557507429
warmup_ratio	0.07903396276412193