metadata

language:
  - en
library_name: span-marker
tags:
  - span-marker
  - token-classification
  - ner
  - named-entity-recognition
  - generated_from_span_marker_trainer
datasets:
  - tomaarsen/ner-orgs
metrics:
  - precision
  - recall
  - f1
widget:
  - text: >-
      Hallacas are also commonly consumed in eastern Cuba parts of Colombia,
      Ecuador, Aruba, and Curaçao.
  - text: >-
      The co-production of Yvon Michel's GYM and Jean Bédard's Interbox
      promotions and televised via HBO, has trumped a proposed HBO -televised
      rematch between Jean Pascal and RING and WBC 175-pound champion Chad
      Dawson that was slated for the same date at Bell Centre in Montreal.
  - text: >-
      The synoptic conditions see a low over southern Norway, bringing warm
      south and southwesterly flows of air up from the inner continental areas
      of Russia and Belarus.
  - text: >-
      The RCIS recommended amongst other things that the Australian Security
      Intelligence Organisation (ASIO) areas of investigation be widened to
      include terrorism.
  - text: >-
      The large network had multiple campuses in Minnesota, Wisconsin, and South
      Dakota.
pipeline_tag: token-classification
co2_eq_emissions:
  emissions: 532.6472478623315
  source: codecarbon
  training_type: fine-tuning
  on_cloud: false
  cpu_model: 13th Gen Intel(R) Core(TM) i7-13700K
  ram_total_size: 31.777088165283203
  hours_used: 3.696
  hardware_used: 1 x NVIDIA GeForce RTX 3090
base_model: bert-base-cased
model-index:
  - name: >-
      SpanMarker with bert-base-cased on FewNERD, CoNLL2003, OntoNotes v5, and
      MultiNERD
    results:
      - task:
          type: token-classification
          name: Named Entity Recognition
        dataset:
          name: FewNERD, CoNLL2003, OntoNotes v5, and MultiNERD
          type: tomaarsen/ner-orgs
          split: test
        metrics:
          - type: f1
            value: 0.8311343653918766
            name: F1
          - type: precision
            value: 0.8334090564894745
            name: Precision
          - type: recall
            value: 0.8288720574945131
            name: Recall

SpanMarker with bert-base-cased on FewNERD, CoNLL2003, OntoNotes v5, and MultiNERD

This is a SpanMarker model trained on the FewNERD, CoNLL2003, OntoNotes v5, and MultiNERD 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: FewNERD, CoNLL2003, OntoNotes v5, and MultiNERD
Language: en

Model Sources

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

Model Labels

Label	Examples
ORG	"IAEA", "Church 's Chicken", "Texas Chicken"

Evaluation

Metrics

Label	Precision	Recall	F1
all	0.8334	0.8289	0.8311
ORG	0.8334	0.8289	0.8311

Uses

Direct Use for Inference

from span_marker import SpanMarkerModel

# Download from the 🤗 Hub
model = SpanMarkerModel.from_pretrained("tomaarsen/span-marker-bert-base-orgs")
# Run inference
entities = model.predict("The large network had multiple campuses in Minnesota, Wisconsin, and South Dakota.")

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-orgs")

# 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-orgs-finetuned")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Sentence length	1	22.1911	267
Entities per sentence	0	0.8144	39

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
0.3273	3000	0.0052
0.6546	6000	0.0047
0.9819	9000	0.0045
1.3092	12000	0.0047
1.6365	15000	0.0045
1.9638	18000	0.0046
2.2911	21000	0.0054
2.6184	24000	0.0053
2.9457	27000	0.0052

Environmental Impact

Carbon emissions were measured using CodeCarbon.

Carbon Emitted: 0.533 kg of CO2
Hours Used: 3.696 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}
}