Developed by Arda Akdemir, as part of the EU-funded Emerging Welfare Work Package 2 (Information Extraction) in Koc University under the supervision of Deniz Yuret.
We made use of the kNet package to develop our model which is a deep learning package also developed in Koc University. The model is the reimplementation of the state-of-the-art named entity recognition model End-to-end Sequence Labeling via Bi-directional LSTM-CNNs-CRF.
main.jl is the main file of the model. Running the model automatically trains a model using the data provided in the data folder.
The dataset we have used during training and testing are included in the repository train.txt, valid.txt and test.txt. The dataset is the splitted version of the CoNLL-2003 Named Entity Recognition dataset.
We have obtained comparable results with the reference paper. Our results are included inside repository.