shreyasseshadri / IR-Project

IR Project for Team 5

IR-Project

There are two drive folders consisting of two implementations

• Supervised- Bidaf link
• Unsuperised- Word2vec+cosine similarity link

Instructions

General Instructions

• This repository contains only the code it cannot can be directly run, to run the code, please use the drive link given above as that folder contains all the required data to run.
• Change notebook settings to use GPU to get maximum performance.

Unsupervised Method

• Locate unsupervised.ipynb in the folder for supervised method implementation of this approach.
• It has three sections, which are
  • Pre Processing
  • Training Word2vec model
  • Cosine Similarity
• All data folders are already downloaded and setup in the folder.

Supervised Bidaf

• Locate main.ipynb in the folder for supervised method(given above) and execute it line by line.
• Installation of required dependencies is already given in the cells, and just needs to be executed.
• It also uses the evaluation script given SqUAD, for evaluation of the predicions.
• Layers folder consists of all the custom built Keras Layers.

Team Members

S.Shreyas - 16IT135
Sanjay Shetty - 16IT136