SiameseCBOW

Implementation of SiameseCBOW using python3.4, keras and tensorflow.

Environments

• Mac OS 10.10.5
• python 3.4.3
• Keras 2.1.2
• tensorflow 1.4.0
  • Mac, CPU version:https://storage.googleapis.com/tensorflow/mac/cpu/tensorflow-1.4.0-py3-none-any.whl

How to use

0. Installation

$ pip install -r requirements.txt

1. Customize src/load.py

At first, you should customize src/load.py so that you can load your data and preprocess it.

If you just want to test the codes, please comment out x, y = load(file) and set variable:x, y in main.py like this:

def main():

...

 import numpy as np
 x = [np.ones((10, input_length))] * (1 + n_positive + n_negative)
 y = np.ones((10, n_positive + n_negative))

...

 model.fit(x, y, epochs=1)

2. Set Hyper-parameters

If you want, please set Hyper-parameters such as embedding dimension in main.py.

3. Train&Save

Please execute this command at the git project directory:

$ python main.py -f <data_path>

and a pickle file of an embedding vector will be saved in ./save/.