Word2Vec Negative_Sampling

This is an efficient implementation of Word2vec on game of thrones textbooks
Note: This implemented on Windows OS, please find all path strings and change \ with / if running on linux or Mac

Dependecies:

1. Tensorflow
2. Python 3
3. Numpy
4. os
5. argpase
6. glob

Train From Scratch

if you would like to run the model yourself and configure the hyper-parameters specified in main.py please do delete the following folders first to avoid conflicts when running tensorflow:

1. visualizations
2. graph
3. checkpoints

To train from scratch have a look at main.py and choose the hyper-parameters you would like to experiment with, there is only 2 mandatory arguments --data-dir and --vocab-dir

python main.py --data-dir data\\ --vocab-dir vocab\\

Just-Visualization

you can use my trained model and run tensorboard to visualize the word vectors generated; to do so:

1- open terminal (cmd on windows) 
2- Navigate to visualizations folder
3- run 
tensorboard --logdir=visualizations`
4- copy and paste the url provided by tensorboard in chrome
5- load the vocab_3000.tsv file located in visualization folder in tensorboard to identify each word

Evaluation

you may as well run evaluate.py to find analogies and nearest words regarding game of thrones
my favourite one is
Mother is to Joffrey as "ghost/Sam" is to Jon