Text Triplet Loss using tensorflow

This repository provides a CNN implementation of Triplet Loss on Text data.

Requirements

1. Python3
2. Tensorflow
3. Numpy
4. Fasttext

Setup

Run setup.sh to setup required library and dependencies.

bash setup.sh

Download pre-trained model from following Google Drive repository.

https://drive.google.com/file/d/1VL6BThgkRzfSjD-WtfukSvKPiwOZ-4a7/view?usp=sharing

Unzip and paste ./model_triplets/ directory to the project root directory.

Pre-Trained Model:

Directory structure for pre-trained model should be as follows:

1. ./model_triplets/checkpoint
2. ./model_triplets/model.ckpt.data-00000-of-00001
3. ./model_triplets/model.ckpt.index
4. ./model_triplets/model.ckpt.meta
5. ./model_triplets/vocab

Training:

Run train_triplets.py to train model.

python train_triplets.py

Prediction:

python predict.py "how to become a data scientist?"

Use Prediction.ipynb to get more detailed prediction code.