RNN and Captioning for Gender Classification

Gender Classification From Tweets and Posted Images

Requirements

Python 2.7 (it may work on python3 as well, not guaranteed)
(Preferable) CUDA 9.0 and Nvidia GPU compatible with CUDA
Word embeddings, we prefer GLoVe (https://nlp.stanford.edu/projects/glove/)
As a corpus to test, we used PAN2018 Author Profiling dataset (https://pan.webis.de/clef18/pan18-web/author-profiling.html)

To change parameters you can use parameters.py file, it has all paths, hyperparameters, dimension fields.
Run.sh make it easy to test several languages and parameters in one big run. After each language Training&Testing it calls model deleter which deletes each saved model but the first 5.
Making Optimize=True means that you are going to test some hyperparameters and save models that got higher than your model_save_threshold. If you make it false you will just train a model without saving any weigths.
To switch into other models that have been tested for Gender Prediction problem, you can switch branches, we hope their names are self-exploratory.

Name		Name	Last commit message	Last commit date
Latest commit History 2 Commits
.idea		.idea
__pycache__		__pycache__
LICENSE		LICENSE
README.md		README.md
eval.py		eval.py
eval.pyc		eval.pyc
main.py		main.py
model.py		model.py
model.pyc		model.pyc
modelDeleter.py		modelDeleter.py
parameters.py		parameters.py
parameters.pyc		parameters.pyc
parameters_ar.py		parameters_ar.py
parameters_en.py		parameters_en.py
parameters_es.py		parameters_es.py
preprocess.py		preprocess.py
preprocess.pyc		preprocess.pyc
run.sh		run.sh
test.py		test.py
train.py		train.py
train.pyc		train.pyc
visualizer.py		visualizer.py