This project aims to solve the task stated in NEEL homepage The aim is to use a Bi-LSTM neural network in order to solve the NER task stated before. The project relies also upon the DBPedia API in order to link the found entities.
To run the project you will need the following setup:
- Python 3.x environment
- Tensorflow 1.2 +
- Python Requests module (
pip3 install requests
) - Fill the dataset (embeddings too) with your own (no working copy of the original dataset is included for license issues)
Run in a terminal:
python3 main.py --pretrain --train_nfold
This will get the model to be pre-trained and cross validated.
To run the interactive shell (the model is the one saved in the trained
folder) you will simply need to run the following:
python3 main.py --interactive
For further info please run the following:
python3 main.py --help
When the train and testing routines are finished, a metrics.txt
file will be produced in the working directory.
Please, use the macros.py
file to edit preferences.
For any issue, please contact me here