stefan-it's picture
readme: add initial version
78aedc2
metadata
language: nl
license: mit
tags:
  - flair
  - token-classification
  - sequence-tagger-model
base_model: hmteams/teams-base-historic-multilingual-discriminator
widget:
  - text: >-
      Professoren der Geneeskun dige Faculteit te Groningen alsook van de HH ,
      Doctoren en Chirurgijns van Groningen , Friesland , Noordholland ,
      Overijssel , Gelderland , Drenthe , in welke Provinciën dit Elixir als
      Medicament voor Mond en Tanden reeds jaren bakend is .

Fine-tuned Flair Model on Dutch ICDAR-Europeana NER Dataset

This Flair model was fine-tuned on the Dutch ICDAR-Europeana NER Dataset using hmTEAMS as backbone LM.

The ICDAR-Europeana NER Dataset is a preprocessed variant of the Europeana NER Corpora for Dutch and French.

The following NEs were annotated: PER, LOC and ORG.

Results

We performed a hyper-parameter search over the following parameters with 5 different seeds per configuration:

  • Batch Sizes: [8, 4]
  • Learning Rates: [3e-05, 5e-05]

And report micro F1-score on development set:

Configuration Run 1 Run 2 Run 3 Run 4 Run 5 Avg.
bs8-e10-lr3e-05 0.8791 0.88 0.8744 0.8843 0.8829 88.01 ± 0.34
bs4-e10-lr3e-05 0.8599 0.8681 0.872 0.8684 0.8851 87.07 ± 0.82
bs8-e10-lr5e-05 0.8688 0.86 0.8726 0.8681 0.8772 86.93 ± 0.57
bs4-e10-lr5e-05 0.8622 0.8684 0.8617 0.8667 0.8651 86.48 ± 0.26

The training log and TensorBoard logs (only for hmByT5 and hmTEAMS based models) are also uploaded to the model hub.

More information about fine-tuning can be found here.

Acknowledgements

We thank Luisa März, Katharina Schmid and Erion Çano for their fruitful discussions about Historic Language Models.

Research supported with Cloud TPUs from Google's TPU Research Cloud (TRC). Many Thanks for providing access to the TPUs ❤️