Setup

The project is based on Deep Learning Neural Network. The code has been tested on Ubuntu 16.04 LTS.

Requirements

• Python 3.3+ or Python 2.7
• macOS or Linux or Windows

Dependencies :

1. Python
2. Keras
3. Pandas
4. Matplotlib
5. Hdf5
6. H5py
7. Nltk.corpus
8. re
9. Pickle
10. Numpy
11. Glove Word2Vec

Instructions:

1. Place glove folder in extracted form in the same directory of codes.

2. Download quora dataset from “http://qim.ec.quoracdn.net/quora_duplicate_questions.tsv”. Place it in the same directory of codes in extracted form. Name of this dataset should be different from “train.csv” and “test.csv”

Run:

1. python make_test_train_data_from_given_quora_dataset <dataset_name.csv>

2. For LSTM with word embedding:

	Training :  python siamese_lstm_word.py train
	
	Testing  :  python siamese_lstm_word.py test

3. For LSTM with char embedding:

  	  python char_embedding.py

          Training :  python siamese_lstm_char.py train
   
          Testing  :  python siamese_lstm_char.py test

4. For BiLSTM with word embedding:

  	Training :  python siamese_BiLSTM_word.py train
	
	Testing  :  python siamese_BiLSTM_word.py test

Documentation

The documentation of the latest released version of Duplicate Question Pair Detection is available here.