tomaarsen/span-marker-bert-base-fewnerd-coarse-super

SpanMarker with bert-base-cased on DFKI-SLT/few-nerd

This is a SpanMarker model trained on the DFKI-SLT/few-nerd dataset that can be used for Named Entity Recognition. This SpanMarker model uses bert-base-cased as the underlying encoder.

Model Details

Model Description

• Model Type: SpanMarker
• Encoder: bert-base-cased
• Maximum Sequence Length: 256 tokens
• Maximum Entity Length: 8 words
• Training Dataset: DFKI-SLT/few-nerd
• Language: en
• License: cc-by-sa-4.0

Model Sources

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

Model Labels

Label	Examples
art	"Imelda de ' Lambertazzi", "Time", "The Seven Year Itch"
building	"Sheremetyevo International Airport", "Henry Ford Museum", "Boston Garden"
event	"Russian Revolution", "Iranian Constitutional Revolution", "French Revolution"
location	"Croatian", "Mediterranean Basin", "the Republic of Croatia"
organization	"Texas Chicken", "IAEA", "Church 's Chicken"
other	"Amphiphysin", "N-terminal lipid", "BAR"
person	"Hicks", "Edmund Payne", "Ellaline Terriss"
product	"Corvettes - GT1 C6R", "Phantom", "100EX"

Evaluation

Metrics

Label	Precision	Recall	F1
all	0.7767	0.7697	0.7732
art	0.7419	0.7488	0.7454
building	0.6237	0.7102	0.6641
event	0.6154	0.5530	0.5825
location	0.8122	0.8515	0.8314
organization	0.7321	0.6898	0.7103
other	0.7376	0.6329	0.6812
person	0.8805	0.9062	0.8932
product	0.6641	0.5898	0.6248

Uses

Direct Use for Inference

from span_marker import SpanMarkerModel

# Download from the 🤗 Hub
model = SpanMarkerModel.from_pretrained("span_marker_model_id")
# Run inference
entities = model.predict("Caretaker manager George Goss led them on a run in the FA Cup , defeating Liverpool in round 4 , to reach the semi-final at Stamford Bridge , where they were defeated 2–0 by Sheffield United on 28 March 1925 .")

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("span_marker_model_id")

# 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("span_marker_model_id-finetuned")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Sentence length	1	24.4956	163
Entities per sentence	0	2.5439	35

Training Hyperparameters

• learning_rate: 5e-05
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 1

Training Results

Epoch	Step	Validation Loss	Validation Precision	Validation Recall	Validation F1	Validation Accuracy
0.1629	200	0.0318	0.6968	0.5990	0.6442	0.8994
0.3259	400	0.0239	0.7350	0.7187	0.7268	0.9227
0.4888	600	0.0216	0.7628	0.7205	0.7410	0.9294
0.6517	800	0.0206	0.7681	0.7573	0.7626	0.9338
0.8147	1000	0.0196	0.7743	0.7540	0.7640	0.9358
0.9776	1200	0.0193	0.7719	0.7673	0.7696	0.9378

Environmental Impact

Carbon emissions were measured using CodeCarbon.

• Carbon Emitted: 0.015 kg of CO2
• Hours Used: 0.14 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: 1.3.1.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}
}