dalmia / Quora-Question-Pairs

The code for our submission in Kaggle's competition Quora Question Pairs which ranked in the top 25%.

Quora-Question-Pairs

This repository contains the code for our submission in Kaggle's competition Quora Question Pairs in which we ranked in the top 25%. A detailed report for the project can be found here.

Data

train.csv contains ~ 400k question pairs along with the corresponding label (duplicate or not) and test.csv contains ~ 2300k question pairs. Both the files can be found here.

Model Architecture

We use a Siamese Neural Network architecture with Gated Recurrent Units in combination with traditional Machine Learning algorithms like Random Forest, SVM and Adaboost.

Running the model

Firstly, place the train.csv, test.csv (see the Data section above) and the pre-trained GloVe embeddings in the input folder. You can download the embeddings from here. Then, simply run the bash script:

bash run_model.sh

Contributors

• Ameya Godbole
• Aman Dalmia

Dependencies

• numpy
• pandas
• nltk
• sklearn
• TensorFlow

Install them using pip.

Note

• If there is any issue running the code, please post it in the issue tracker.
• If you like this repo and find it useful, please consider ★ starring it (on top right of the page) :)