SpanMarker with bert-base-multilingual-cased on TLUnified

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

Model Details

Model Description

Model Type: SpanMarker
Encoder: bert-base-multilingual-cased
Maximum Sequence Length: 256 tokens
Maximum Entity Length: 8 words
Training Dataset: TLUnified
Language: tl
License: gpl-3.0

Model Sources

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

Model Labels

Label	Examples
LOC	"Israel", "Batasan", "United States"
ORG	"MMDA", "International Monitoring Team", "Coordinating Committees for the Cessation of Hostilities"
PER	"Puno", "Fernando", "Villavicencio"

Evaluation

Metrics

Label	Precision	Recall	F1
all	0.8737	0.9042	0.8887
LOC	0.8830	0.9084	0.8955
ORG	0.7579	0.8587	0.8052
PER	0.9264	0.9220	0.9242

Uses

Direct Use for Inference

from span_marker import SpanMarkerModel

# Download from the 🤗 Hub
model = SpanMarkerModel.from_pretrained("tomaarsen/span-marker-mbert-base-tlunified")
# Run inference
entities = model.predict("Idinagdag ni South Cotabato Rep Darlene Antonino - Custodio, na illegal na ipagpaliban ang halalan sa ARMM kung ang gagamitin lamang basehan ay ang ipapasang panukala ng Kongreso.")

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-mbert-base-tlunified")

# 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-mbert-base-tlunified-finetuned")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Sentence length	1	31.7625	150
Entities per sentence	0	2.0661	38

Training Hyperparameters

learning_rate: 5e-05
train_batch_size: 16
eval_batch_size: 16
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
0.6803	400	0.0074	0.8552	0.8835	0.8691	0.9774
1.3605	800	0.0072	0.8709	0.9034	0.8869	0.9798
2.0408	1200	0.0070	0.8753	0.9053	0.8900	0.9812
2.7211	1600	0.0065	0.8876	0.9003	0.8939	0.9807

Environmental Impact

Carbon emissions were measured using CodeCarbon.

Carbon Emitted: 0.022 kg of CO2
Hours Used: 0.238 hours

Training Hardware

On Cloud: No
GPU Model: 1 x NVIDIA GeForce RTX 3090
CPU Model: 13th Gen Intel(R) Core(TM) i7-13700K
RAM Size: 31.78 GB

Framework Versions

Python: 3.9.16
SpanMarker: 1.5.1.dev
Transformers: 4.30.0
PyTorch: 2.0.1+cu118
Datasets: 2.14.0
Tokenizers: 0.13.3

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-mbert-base-tlunified