davanstrien/numind_generic-entity_recognition_NER-multilingual-v1_wikiann_de

SpanMarker with numind/generic-entity_recognition_NER-multilingual-v1 on wikiann

This is a SpanMarker model trained on the wikiann dataset that can be used for Named Entity Recognition. This SpanMarker model uses numind/generic-entity_recognition_NER-multilingual-v1 as the underlying encoder.

Model Details

Model Description

• Model Type: SpanMarker
• Encoder: numind/generic-entity_recognition_NER-multilingual-v1
• Maximum Sequence Length: 256 tokens
• Maximum Entity Length: 9 words
• Training Dataset: wikiann
• Language: de
• License: mit

Model Sources

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

Model Labels

Label	Examples
LOC	"Savoyer Voralpen", "Bagan", "Zechin"
ORG	"NHL Entry Draft", "SKA Sankt Petersburg", "Minnesota Wild"
PER	"Antonina Wladimirowna Kriwoschapka", "Lou Salomé", "Jaan Kirsipuu"

Evaluation

Metrics

Label	Precision	Recall	F1
all	0.9070	0.9070	0.9070
LOC	0.9036	0.9298	0.9165
ORG	0.8638	0.8446	0.8541
PER	0.9507	0.9405	0.9455

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("Sein Bundesliga-Debüt gab der Angreifer am 23.")

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	9.7693	85
Entities per sentence	1	1.3821	20

Training Hyperparameters

• learning_rate: 5e-05
• train_batch_size: 64
• eval_batch_size: 128
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 128
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 10
• mixed_precision_training: Native AMP

Training Results

Epoch	Step	Validation Loss	Validation Precision	Validation Recall	Validation F1	Validation Accuracy
1.2658	200	0.0172	0.8842	0.8534	0.8686	0.9586
2.5316	400	0.0145	0.8977	0.8889	0.8933	0.9670
3.7975	600	0.0161	0.8962	0.9006	0.8984	0.9688
5.0633	800	0.0180	0.8982	0.8996	0.8989	0.9689
6.3291	1000	0.0201	0.9014	0.9008	0.9011	0.9694
7.5949	1200	0.0201	0.9010	0.9057	0.9033	0.9702
8.8608	1400	0.0217	0.9062	0.9036	0.9049	0.9702

Framework Versions

• Python: 3.10.12
• SpanMarker: 1.5.0
• Transformers: 4.35.2
• PyTorch: 2.1.0+cu118
• Datasets: 2.15.0
• Tokenizers: 0.15.0

Citation

BibTeX

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