SpanMarker with bert-base-uncased on SourceData

This is a SpanMarker model trained on the SourceData dataset that can be used for Named Entity Recognition. This SpanMarker model uses bert-base-uncased as the underlying encoder.

Model Details

Model Description

Model Type: SpanMarker
Encoder: bert-base-uncased
Maximum Sequence Length: 256 tokens
Maximum Entity Length: 8 words
Training Dataset: SourceData
Language: en
License: cc-by-4.0

Model Sources

Repository: SpanMarker on GitHub
Thesis: SpanMarker For Named Entity Recognition

Model Labels

Label	Examples
CELL_LINE	"293T", "WM266.4 451Lu", "501mel"
CELL_TYPE	"BMDMs", "protoplasts", "epithelial"
DISEASE	"melanoma", "lung metastasis", "breast prostate cancer"
EXP_ASSAY	"interactions", "Yeast two-hybrid", "BiFC"
GENEPROD	"CPL1", "FREE1 CPL1", "FREE1"
ORGANISM	"Arabidopsis", "yeast", "seedlings"
SMALL_MOLECULE	"polyacrylamide", "CHX", "SDS polyacrylamide"
SUBCELLULAR	"proteasome", "D-bodies", "plasma"
TISSUE	"Colon", "roots", "serum"

Evaluation

Metrics

Label	Precision	Recall	F1
all	0.8345	0.8328	0.8336
CELL_LINE	0.9060	0.8866	0.8962
CELL_TYPE	0.7365	0.7746	0.7551
DISEASE	0.6204	0.6531	0.6363
EXP_ASSAY	0.7224	0.7096	0.7160
GENEPROD	0.8944	0.8960	0.8952
ORGANISM	0.8752	0.8902	0.8826
SMALL_MOLECULE	0.8304	0.8223	0.8263
SUBCELLULAR	0.7859	0.7699	0.7778
TISSUE	0.8134	0.8056	0.8094

Uses

Direct Use for Inference

from span_marker import SpanMarkerModel

# Download from the 🤗 Hub
model = SpanMarkerModel.from_pretrained("tomaarsen/span-marker-bert-base-uncased-sourcedata")
# Run inference
entities = model.predict("Comparison of ENCC-derived neurospheres treated with intestinal extract from hypoganglionosis rats, hypoganglionosis treated with Fecal microbiota transplantation (FMT) sham rat. Comparison of neuronal markers. (J) Immunofluorescence stain number of PGP9.5+. Nuclei were stained blue with DAPI; Triangles indicate PGP9.5+.")

Downstream Use

You can finetune this model on your own dataset.

Click to expand

from span_marker import SpanMarkerModel, Trainer

# Download from the 🤗 Hub
model = SpanMarkerModel.from_pretrained("tomaarsen/span-marker-bert-base-uncased-sourcedata")

# Specify a Dataset with "tokens" and "ner_tag" columns
dataset = load_dataset("conll2003") # For example CoNLL2003

# Initialize a Trainer using the pretrained model & dataset
trainer = Trainer(
    model=model,
    train_dataset=dataset["train"],
    eval_dataset=dataset["validation"],
)
trainer.train()
trainer.save_model("tomaarsen/span-marker-bert-base-uncased-sourcedata-finetuned")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Sentence length	4	71.0253	2609
Entities per sentence	0	8.3186	162

Training Hyperparameters

learning_rate: 5e-05
train_batch_size: 32
eval_batch_size: 32
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_ratio: 0.1
num_epochs: 3

Training Results

Epoch	Step	Validation Loss	Validation Precision	Validation Recall	Validation F1	Validation Accuracy
0.5237	3000	0.0162	0.7972	0.8162	0.8065	0.9520
1.0473	6000	0.0155	0.8188	0.8251	0.8219	0.9560
1.5710	9000	0.0155	0.8213	0.8324	0.8268	0.9563
2.0946	12000	0.0163	0.8315	0.8347	0.8331	0.9581
2.6183	15000	0.0167	0.8303	0.8378	0.8340	0.9582

Framework Versions

Python: 3.9.16
SpanMarker: 1.3.1.dev
Transformers: 4.33.0
PyTorch: 2.0.1+cu118
Datasets: 2.14.0
Tokenizers: 0.13.2

Citation

BibTeX

@software{Aarsen_SpanMarker,
    author = {Aarsen, Tom},
    license = {Apache-2.0},
    title = {{SpanMarker for Named Entity Recognition}},
    url = {https://github.com/tomaarsen/SpanMarkerNER}
}

tomaarsen
/

span-marker-bert-base-uncased-sourcedata