Test your English language skills!

Puffin is a Web App with two games BERTMatch and BERTQuiz. It uses a pretrained neural network to compute semantic similarity and evaluate user response. BERTQuiz is a Reading Comprehension game and BERTMatch checks how well one can paraphrase a sentence.

• Python, Flask, Tensorflow, etc as listed in requirements.txt

ress the Fork button (on the top right corner of the page) to save a copy of this project on your account. Download the repository as follows:

git clone https://github.com/sramakrishnan247/Puffin

Python and its dependencies

apt-get install -y python python-setuptools python-dev build-essential python-pip 

Install Python Flask and Tensorflow dependencies

pip install -r requirements.txt

Start web server

FLASK_APP=app.py flask run --host=0.0.0.0

Open a browser and go to URL

http://localhost:5000   
Enjoy!

We build and experiment with different deep learning models to predict the semantic similarity between two sentences. From our experiments, we observed that the transformer based BERT models outperformed the LSTM network and also trained faster. Moreover, the BERT models that were finetuned with a Bi-directional LSTM gave a much better performance than the BERT model that was not finetuned to the specific task of semantic similarity recognition. We have developed two games, BertMatch(Paraphrase a sentence) and BertQuiz(Reading Comprehension), both of which are fun to play and a good way to learn English language!