TELEGRAM DATA CLUSTERING CONTEST

Overview

This repo is our attemp on the Telegram Data Clustering contest. More details here : https://contest.com/docs/data_clustering

Setting up:

1. The submission folder contains the files that will be finally shared for submission.
2. Install the dependencies in deb_packages.txt
3. site_packages are prepared to have the correct dependencies for the debian system that will be used for the evaluation as mentioned in the contest page.
4. You can create a virtualenv and install the pythin dependencies listed in requirements.txt by doing pip install -r requirements.txt
5. Additionally, some pre-trained vectors need to the downloaded and added to the assets folders.

Execution

1. tgnews executable has been set up specifically for the contest.
2. For testing, run the modules using

   (venv) submission :  python src/main.py languages 'test/'
   
   

   and so on.

Explorations

1. Jupyter notebooks show the different EDA, trials and modelling section.