Content Summarization

Inspiration

The desire to work on a machine-learning problem using game-changing tools to create a microservice.

What it does

The service is a text summarisation service that takes text input and comes out with its summary. It is an extractive summariser with plans underway for more abstractive summaries.

How we built it

We used BERT embeddings with a clustering approach all built on pytorch and python with a react based frontend.

Challenges we ran into

Leveraging the BERT embeddings to give results faster was a challenge.

Accomplishments that we're proud of

A faster microservice An amalgamation of BERT embeddings and the clustering technique.

What we learned

We learnt the mechanism of BERT and its embeddings, KMeans and its intricacies.

What's next for Content Summarization

Better summarisation that could be done by using an ensemble model of different strategies plus using abstractive summarisation.

Built With

• azure
• bert
• fastapi
• javascript
• mongodb
• python
• pytorch
• react