Hugging Face's logo

JohanHeinsen
/
PE_efterlyst_classifier_v2

Text Classification
setfit
Safetensors
sentence-transformers
Danish
bert
generated_from_setfit_trainer
Eval Results (legacy)
text-embeddings-inference
Model card Files
xet
Community

SetFit with JohanHeinsen/Old_News_Segmentation_SBERT_V0.1

This is a SetFit model that can be used for Text Classification. This SetFit model uses JohanHeinsen/Old_News_Segmentation_SBERT_V0.1 as the Sentence Transformer embedding model. A LogisticRegression instance is used for classification. It is designed to identify texts describing missing people from police gazettes in nineteenth century Denmark.

The model has been trained using an efficient few-shot learning technique that involves:

  1. Fine-tuning a Sentence Transformer with contrastive learning.
  2. Training a classification head with features from the fine-tuned Sentence Transformer.

Model Details

Model Description

Model Sources

Model Labels

Label Examples
1
  • '2) Fattiglem, fhv. Roersbetjent, Jens Hansen, af Fare, har den 15de ds. forladt Fattiggaarden. der og formodes at drive arbejdsløs omkring muligvis er han taget til Kjøbenhavn for at søge Hyre som Sømand. Saafremt han, der er 50 Aar gl., middel af Væxt, og var iført sort rundpullet Hat, blaa Trøje og Benklæder muligvis dog et Par Lærreds ovenpaa – samt Træsko, maatte antræffes, bedes Underretning derom meddelt Byog Herredsfogden i Storehedinge.'
  • '1) En Mandsperson, ca. 20 Aar gl., middelstor eller lidt mindre, lyst Polkahaar, ordentlig klædt med mørk Sækfrakke og lyse Buxer, – sigtes for Tyveriet Nr. 3. (St. 2, 291.)'
  • '2) Oplysning om, hvor Garversvend Niels Peter Schmidt eller Niels Peter Nielsen Schmidt, født 16de Febr. 1839, maatte opholde sig, bedes meddeelt Muckadell m. fl. Birkers Kontor i Spanget pr. Kværndrup. Paagjældende blev blev den 28de Januar d. A. viseret derfra til Odense, hvorfra han strax igien skal være afgaaet til Fredericia.'
0
  • '2) 2 Høns og en Hane, denne sidste graa med laadne Ben, den ene Høne brunspættet, den anden sort med laadne Ben, ere bortkomne siden den 24. f.M. (St. 7, 448).'
  • 'Hans Edvard Valdemar Holst (Kbhvn.), 45 Aar. Løsgængeri.'
  • 'Peter Christian Leyring (Levring), 26 Aar. Betleri.'

Evaluation

Metrics

Label Accuracy F1 Precision Recall
all 0.9817 0.9385 0.9231 0.9545

Training Details

Training Set Metrics

Training set Min Median Max
Word count 3 20.8907 245
Label Training Sample Count
0 1195
1 205

Training Hyperparameters

  • batch_size: (16, 16)
  • num_epochs: (3, 3)
  • max_steps: -1
  • sampling_strategy: oversampling
  • num_iterations: 12
  • body_learning_rate: (2e-05, 2e-05)
  • head_learning_rate: 2e-05
  • loss: CosineSimilarityLoss
  • distance_metric: cosine_distance
  • margin: 0.25
  • end_to_end: False
  • use_amp: False
  • warmup_proportion: 0.1
  • l2_weight: 0.01
  • seed: 87
  • eval_max_steps: -1
  • load_best_model_at_end: False

Training Results

Epoch Step Training Loss Validation Loss
0.0005 1 0.1797 -
0.0238 50 0.2091 -
0.0476 100 0.1061 -
0.0714 150 0.0529 -
0.0952 200 0.0491 -
0.1190 250 0.0238 -
0.1429 300 0.0195 -
0.1667 350 0.013 -
0.1905 400 0.0066 -
0.2143 450 0.005 -
0.2381 500 0.0038 -
0.2619 550 0.0038 -
0.2857 600 0.005 -
0.3095 650 0.0062 -
0.3333 700 0.0024 -
0.3571 750 0.0002 -
0.3810 800 0.0003 -
0.4048 850 0.0008 -
0.4286 900 0.0001 -
0.4524 950 0.0006 -
0.4762 1000 0.0022 -
0.5 1050 0.0003 -
0.5238 1100 0.0016 -
0.5476 1150 0.0001 -
0.5714 1200 0.0 -
0.5952 1250 0.0 -
0.6190 1300 0.0 -
0.6429 1350 0.0 -
0.6667 1400 0.0 -
0.6905 1450 0.0 -
0.7143 1500 0.0 -
0.7381 1550 0.0024 -
0.7619 1600 0.0002 -
0.7857 1650 0.0001 -
0.8095 1700 0.0 -
0.8333 1750 0.0 -
0.8571 1800 0.0 -
0.8810 1850 0.0 -
0.9048 1900 0.0 -
0.9286 1950 0.0 -
0.9524 2000 0.0 -
0.9762 2050 0.0 -
1.0 2100 0.0 -
1.0238 2150 0.0 -
1.0476 2200 0.0 -
1.0714 2250 0.0 -
1.0952 2300 0.0 -
1.1190 2350 0.0 -
1.1429 2400 0.0 -
1.1667 2450 0.0 -
1.1905 2500 0.0 -
1.2143 2550 0.0 -
1.2381 2600 0.0 -
1.2619 2650 0.0 -
1.2857 2700 0.0 -
1.3095 2750 0.0 -
1.3333 2800 0.0 -
1.3571 2850 0.0 -
1.3810 2900 0.0 -
1.4048 2950 0.0 -
1.4286 3000 0.0 -
1.4524 3050 0.0 -
1.4762 3100 0.0 -
1.5 3150 0.0 -
1.5238 3200 0.0 -
1.5476 3250 0.0 -
1.5714 3300 0.0 -
1.5952 3350 0.0 -
1.6190 3400 0.0 -
1.6429 3450 0.0 -
1.6667 3500 0.0 -
1.6905 3550 0.0 -
1.7143 3600 0.0 -
1.7381 3650 0.0 -
1.7619 3700 0.0 -
1.7857 3750 0.0 -
1.8095 3800 0.0 -
1.8333 3850 0.0 -
1.8571 3900 0.0 -
1.8810 3950 0.0 -
1.9048 4000 0.0 -
1.9286 4050 0.0 -
1.9524 4100 0.0 -
1.9762 4150 0.0 -
2.0 4200 0.0 -
2.0238 4250 0.0 -
2.0476 4300 0.0 -
2.0714 4350 0.0 -
2.0952 4400 0.0 -
2.1190 4450 0.0 -
2.1429 4500 0.0 -
2.1667 4550 0.0 -
2.1905 4600 0.0 -
2.2143 4650 0.0 -
2.2381 4700 0.0 -
2.2619 4750 0.0 -
2.2857 4800 0.0 -
2.3095 4850 0.0 -
2.3333 4900 0.0 -
2.3571 4950 0.0 -
2.3810 5000 0.0 -
2.4048 5050 0.0 -
2.4286 5100 0.0 -
2.4524 5150 0.0 -
2.4762 5200 0.0 -
2.5 5250 0.0 -
2.5238 5300 0.0 -
2.5476 5350 0.0 -
2.5714 5400 0.0 -
2.5952 5450 0.0 -
2.6190 5500 0.0 -
2.6429 5550 0.0 -
2.6667 5600 0.0 -
2.6905 5650 0.0 -
2.7143 5700 0.0 -
2.7381 5750 0.0 -
2.7619 5800 0.0 -
2.7857 5850 0.0 -
2.8095 5900 0.0 -
2.8333 5950 0.0 -
2.8571 6000 0.0 -
2.8810 6050 0.0 -
2.9048 6100 0.0 -
2.9286 6150 0.0 -
2.9524 6200 0.0 -
2.9762 6250 0.0 -
3.0 6300 0.0 -

Framework Versions

  • Python: 3.11.12
  • SetFit: 1.1.3
  • Sentence Transformers: 4.1.0
  • Transformers: 4.51.3
  • PyTorch: 2.7.0
  • Datasets: 2.19.2
  • Tokenizers: 0.21.1
Downloads last month
1
Safetensors
Model size
0.1B params
Tensor type
F32
·
Inference Providers NEW
Text Classification
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for JohanHeinsen/PE_efterlyst_classifier_v2

Base model

CALDISS-AAU/DA-BERT_Old_News_V1
Finetuned
JohanHeinsen/Old_News_Segmentation_SBERT_V0.1
Finetuned
(14)
this model

Paper for JohanHeinsen/PE_efterlyst_classifier_v2

Evaluation results