metadata

language: es
license: cc-by-4.0
library_name: span-marker
tags:
  - span-marker
  - token-classification
  - ner
  - named-entity-recognition
  - generated_from_span_marker_trainer
datasets:
  - conll2002
metrics:
  - precision
  - recall
  - f1
widget:
  - text: George Washington fue a Washington.
pipeline_tag: token-classification
base_model: xlm-roberta-large
model-index:
  - name: SpanMarker with xlm-roberta-large on conll2002
    results:
      - task:
          type: token-classification
          name: Named Entity Recognition
        dataset:
          name: conll2002
          type: unknown
          split: eval
        metrics:
          - type: f1
            value: 0.8911398300151355
            name: F1
          - type: precision
            value: 0.8981459751232105
            name: Precision
          - type: recall
            value: 0.8842421441774492
            name: Recall

SpanMarker with xlm-roberta-large on conll2002

This is a SpanMarker model that can be used for Named Entity Recognition. This SpanMarker model uses xlm-roberta-large as the underlying encoder.

Model Details

Model Description

Model Type: SpanMarker
Encoder: xlm-roberta-large
Maximum Sequence Length: 256 tokens
Maximum Entity Length: 8 words
Training Dataset: conll2002
Language: es
License: cc-by-4.0

Model Sources

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

Model Labels

Label	Examples
LOC	"Melbourne", "Australia", "Victoria"
MISC	"CrimeNet", "Ciudad", "Ley"
ORG	"Commonwealth", "Tribunal Supremo", "EFE"
PER	"Abogado General del Estado", "Daryl Williams", "Abogado General"

Uses

Direct Use for Inference

from span_marker import SpanMarkerModel

# Download from the 🤗 Hub
model = SpanMarkerModel.from_pretrained("alvarobartt/span-marker-xlm-roberta-large-conll-2002-es")
# Run inference
entities = model.predict("George Washington fue a Washington.")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Sentence length	1	31.8052	1238
Entities per sentence	0	2.2586	160

Training Hyperparameters

learning_rate: 1e-05
train_batch_size: 16
eval_batch_size: 8
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: 2

Training Results

Epoch	Step	Validation Loss	Validation Precision	Validation Recall	Validation F1	Validation Accuracy
0.0587	50	0.4612	0.0280	0.0007	0.0014	0.8576
0.1174	100	0.0512	0.5	0.0002	0.0005	0.8609
0.1761	150	0.0254	0.7622	0.5494	0.6386	0.9278
0.2347	200	0.0177	0.7840	0.7135	0.7471	0.9483
0.2934	250	0.0153	0.8072	0.7944	0.8007	0.9662
0.3521	300	0.0175	0.8439	0.7544	0.7966	0.9611
0.4108	350	0.0103	0.8828	0.8108	0.8452	0.9687
0.4695	400	0.0105	0.8674	0.8433	0.8552	0.9724
0.5282	450	0.0098	0.8651	0.8477	0.8563	0.9745
0.5869	500	0.0092	0.8634	0.8306	0.8467	0.9736
0.6455	550	0.0106	0.8556	0.8581	0.8568	0.9758
0.7042	600	0.0096	0.8712	0.8521	0.8616	0.9733
0.7629	650	0.0090	0.8791	0.8420	0.8601	0.9740
0.8216	700	0.0082	0.8883	0.8799	0.8840	0.9769
0.8803	750	0.0081	0.8877	0.8604	0.8739	0.9763
0.9390	800	0.0087	0.8785	0.8738	0.8762	0.9763
0.9977	850	0.0084	0.8777	0.8653	0.8714	0.9767
1.0563	900	0.0081	0.8894	0.8713	0.8803	0.9767
1.1150	950	0.0078	0.8944	0.8708	0.8825	0.9768
1.1737	1000	0.0079	0.8973	0.8722	0.8846	0.9776
1.2324	1050	0.0080	0.8792	0.8780	0.8786	0.9783
1.2911	1100	0.0082	0.8821	0.8574	0.8696	0.9767
1.3498	1150	0.0075	0.8928	0.8697	0.8811	0.9774
1.4085	1200	0.0076	0.8919	0.8803	0.8860	0.9792
1.4671	1250	0.0078	0.8846	0.8695	0.8770	0.9781
1.5258	1300	0.0074	0.8944	0.8845	0.8894	0.9792
1.5845	1350	0.0076	0.8922	0.8856	0.8889	0.9796
1.6432	1400	0.0072	0.9004	0.8799	0.8900	0.9790
1.7019	1450	0.0076	0.8944	0.8889	0.8916	0.9800
1.7606	1500	0.0074	0.8962	0.8861	0.8911	0.9800
1.8192	1550	0.0072	0.8988	0.8886	0.8937	0.9809
1.8779	1600	0.0074	0.8962	0.8833	0.8897	0.9797
1.9366	1650	0.0071	0.8976	0.8849	0.8912	0.9799
1.9953	1700	0.0071	0.8981	0.8842	0.8911	0.9799

Framework Versions

Python: 3.10.12
SpanMarker: 1.3.1.dev
Transformers: 4.33.2
PyTorch: 2.0.1+cu118
Datasets: 2.14.5
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}
}