We present community-driven BERT and ELECTRA models for Turkish 🎉

Some datasets used for pretraining and evaluation are contributed from the awesome Turkish NLP community, as well as the decision for the BERT model name: BERTurk.

• 12.05.2020: Release of ELECTRA (small and base) models, see here.
• 25.03.2020: Release of BERTurk uncased model and BERTurk models with larger vocab size (128k, cased and uncased).
• 11.03.2020: Release of the cased distilled BERTurk model: DistilBERTurk. Available on the Hugging Face model hub
• 17.02.2020: Release of the cased BERTurk model. Available on the Hugging Face model hub
• 10.02.2020: Training corpus update, new TensorBoard links, new results for cased model.
• 02.02.2020: Initial version of this repo.

The current version of the model is trained on a filtered and sentence segmented version of the Turkish OSCAR corpus, a recent Wikipedia dump, various OPUS corpora and a special corpus provided by Kemal Oflazer.

The final training corpus has a size of 35GB and 44,04,976,662 tokens.

Thanks to Google's TensorFlow Research Cloud (TFRC) we can train both cased and uncased models on a TPU v3-8. You can find the TensorBoard outputs for the training here:

• TensorBoard cased model
• TensorBoard uncased model

We also provide cased and uncased models that aŕe using a larger vocab size (128k instead of 32k).

A detailed cheatsheet of how the models were trained, can be found here.

The distilled version of a cased model, so called DistilBERTurk, was trained on 7GB of the original training data, using the cased version of BERTurk as teacher model.

DistilBERTurk was trained with the official Hugging Face implementation from here.

The cased model was trained for 5 days on 4 RTX 2080 TI.

More details about distillation can be found in the "DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter" paper by Sanh et al. (2019).

In addition to the BERTurk models, we also trained ELECTRA small and base models. A detailed overview can be found in the ELECTRA section.

We use the token classification example from 🤗/Transformers for evaluation on both PoS and NER datasets. We report averaged Accuracy (PoS tagging) and F-Score (NER) on 5 runs (initialized with 5 different seeds).

BERTurk and ELECTRA model checkpoint selection: We evaluated 5 different checkpoints for our cased and uncased models based on the development score for PoS tagging and NER. The model with the best results is used for the final and released model.

Evaluation is done with the Hugging Face Transformers library and the token classification example script run_ner.py. We use the following hyper-parameters:

The Turkish IMST dataset from Universal Dependencies is used for PoS tagging evaluation. We use the dev branch and commit a6c955. Result on development set is reported in brackets.

NER dataset is similar to the one used in this paper. We converted the dataset into CoNLL-like format and used a 80/10/10 training, development and test split. Result on development set is reported in brackets.

All trained models can be used from the DBMDZ Hugging Face model hub page using their model name. The following models are available:

• BERTurk models with 32k vocabulary: dbmdz/bert-base-turkish-cased and dbmdz/bert-base-turkish-uncased
• BERTurk models with 128k vocabulary: dbmdz/bert-base-turkish-128k-cased and dbmdz/bert-base-turkish-128k-uncased
• ELECTRA small and base cased models (discriminator): dbmdz/electra-small-turkish-cased-discriminator and dbmdz/electra-base-turkish-cased-discriminator

Example usage with 🤗/Transformers:

tokenizer = AutoTokenizer.from_pretrained("dbmdz/bert-base-turkish-cased")

model = AutoModel.from_pretrained("dbmdz/bert-base-turkish-cased")

This loads the BERTurk cased model. The recently introduced ELECTRA base model can be loaded with:

tokenizer = AutoTokenizer.from_pretrained("dbmdz/electra-base-turkish-cased-discriminator")

model = AutoModelWithLMHead.from_pretrained("dbmdz/electra-base-turkish-cased-discriminator")

You can use the following BibTeX entry for citation:

@software{stefan_schweter_2020_3770924,
  author       = {Stefan Schweter},
  title        = {BERTurk - BERT models for Turkish},
  month        = apr,
  year         = 2020,
  publisher    = {Zenodo},
  version      = {1.0.0},
  doi          = {10.5281/zenodo.3770924},
  url          = {https://doi.org/10.5281/zenodo.3770924}
}

Thanks to Kemal Oflazer for providing us additional large corpora for Turkish. Many thanks to Reyyan Yeniterzi for providing us the Turkish NER dataset for evaluation.

Research supported with Cloud TPUs from Google's TensorFlow Research Cloud (TFRC). Thanks for providing access to the TFRC ❤️