metadata

base_model: roberta-base
datasets:
  - conll2003
language:
  - en
library_name: span-marker
license: apache-2.0
metrics:
  - precision
  - recall
  - f1
pipeline_tag: token-classification
tags:
  - span-marker
  - token-classification
  - ner
  - named-entity-recognition
  - generated_from_span_marker_trainer
widget:
  - text: >-
      " The worst thing that could happen for financial markets is that if
      Clinton and Dole start to trade shots in the middle of the ring with
      one-upmanship, " said Hugh Johnson, chief investment officer at First
      Albany Corp. " That's when Wall Street will need to worry . "
  - text: >-
      Poland revived diplomatic ties at ambassadorial level with Yugoslavia in
      April but economic links are almost moribund, despite the end of a
      three-year U.N. trade embargo imposed to punish Belgrade for its support
      of Bosnian Serbs.
  - text: >-
      " We believe that the Israeli settlement policy in the occupied areas is
      an obstacle to the establishment of peace, " German Foreign Ministry
      spokesman Martin Erdmann said.
  - text: >-
      U.S. Agriculture Department officials said Friday that Mexican
      avocados--which are restricted from entering the continental United
      States--will not likely be entering U.S. markets any time soon, even if
      the controversial ban were lifted today.
  - text: 3. Tristan Hoffman (Netherlands) TVM same time
model-index:
  - name: SpanMarker with roberta-base on conll2003
    results:
      - task:
          type: token-classification
          name: Named Entity Recognition
        dataset:
          name: Unknown
          type: conll2003
          split: test
        metrics:
          - type: f1
            value: 0.9022464022464022
            name: F1
          - type: precision
            value: 0.8943980514961726
            name: Precision
          - type: recall
            value: 0.9102337110481586
            name: Recall

SpanMarker with roberta-base on conll2003

This is a SpanMarker model trained on the conll2003 dataset that can be used for Named Entity Recognition. This SpanMarker model uses roberta-base as the underlying encoder.

Model Details

Model Description

Model Type: SpanMarker
Encoder: roberta-base
Maximum Sequence Length: 256 tokens
Maximum Entity Length: 6 words
Training Dataset: conll2003
Language: en
License: apache-2.0

Model Sources

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

Model Labels

Label	Examples
LOC	"BRUSSELS", "Britain", "Germany"
MISC	"British", "EU-wide", "German"
ORG	"EU", "European Commission", "European Union"
PER	"Werner Zwingmann", "Nikolaus van der Pas", "Peter Blackburn"

Evaluation

Metrics

Label	Precision	Recall	F1
all	0.8944	0.9102	0.9022
LOC	0.9220	0.9215	0.9217
MISC	0.7332	0.7949	0.7628
ORG	0.8764	0.8964	0.8863
PER	0.9605	0.9629	0.9617

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("3. Tristan Hoffman (Netherlands) TVM same time")

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	14.5019	113
Entities per sentence	0	1.6736	20

Training Hyperparameters

learning_rate: 1e-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
mixed_precision_training: Native AMP

Training Results

Epoch	Step	Validation Loss	Validation Precision	Validation Recall	Validation F1	Validation Accuracy
0.2775	500	0.0282	0.9105	0.8355	0.8714	0.9670
0.5549	1000	0.0166	0.9215	0.9205	0.9210	0.9824
0.8324	1500	0.0151	0.9247	0.9346	0.9296	0.9853

Framework Versions

Python: 3.10.12
SpanMarker: 1.5.0
Transformers: 4.41.2
PyTorch: 2.3.0+cu121
Datasets: 2.20.0
Tokenizers: 0.19.1

Citation

BibTeX

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