Electra Base Turkish Cased Ner

🚀 Turkish NER Model

This is a Named Entity Recognition (NER) model for Turkish, trained on a specific dataset to identify 48 different entity categories.

🚀 Quick Start

This NER model for Turkish is trained on the Shrinked TWNERTC Turkish NER Data provided by Behçet Şentürk. This dataset is a filtered and cleaned version of the automatically labeled dataset:

Sahin, H. Bahadir; Eren, Mustafa Tolga; Tirkaz, Caglar; Sonmez, Ozan; Yildiz, Eray (2017), “English/Turkish Wikipedia Named-Entity Recognition and Text Categorization Dataset”, Mendeley Data, v1 http://dx.doi.org/10.17632/cdcztymf4k.1

✨ Features

• 48 Categories: Capable of identifying 48 different named entity categories in Turkish text.
• Fine - Tuned: Based on a pre - trained model and fine - tuned for token classification.

🔧 Technical Details

Backbone model

The backbone model is electra-base-turkish-cased-discriminator, which has been fine - tuned for token classification.

The developer is still exploring the possibility of improving the accuracy using the current dataset. However, the model is already suitable for non - critical applications.

Communication and Future Plans

• Discussion: You can reach out to the developer on Twitter for discussions and to report issues.
• Release: A notebook for fine - tuning NER models with Shrinked TWNERTC and sample inference code will be released to show the model's capabilities.