Compendium

Introduction

Compendium is a seq2seq abstractive text symmetrization model based on GRU encoder decoder with attention mechanism.

Base Info

Files and there uses are listed below

Data -> Folder used for train and test data storage
helper -> Contains helper functions used in model.ipynb.
brain -> Contains trained RNN Model.
Data Clean.ipnb -> Ipython Notbook for cleaning and splitting data into train, val and test.
model.ipynb -> Ipython Notebook for training andd testing model.
requirement -> txt file containg required lib

Requirements

Python : 3.X

Pip

Installation

Steps for installation?

Download venv
pip3 install virtualenv
Clone
git clone https://github.com/soni-ratnesh/compendium.git
Change directory
cd compendium
Create and activate virtual environment
pip3 venv env
. env\bin\activate
Install required library
pip3 install -r requirements.txt
Install spacy english model
python3 -m spacy download en
Copy brain file
cp <your barin file path> ./application/model/brain
Run Flask server
bash start.sh

Result

The testing accuracy and loss are,

Test Loss     :  2.23
Test PPE      :  10.87

Need trained model?

We do provide trained model, just let us know. you can get trained model from here . Save the provideed file in brain dir to load and run.

Contribution

Pull requests are welcome. If have an idea please let me know through an issue. For contribution please raise pull requests by,

Clone Repo

git clone https://github.com/soni-ratnesh/compendium.git
Install Dependencies

pip3 install requirements.txt
Verify your changes
Submit Pull Request

Short for time??

Feel free to raise an issue to correct errors or contribute content without a pull request.

djimit/compendium