Machine Learning

Abstract Analyzer

When a paper is being submitted to the arxiv preprint server, the author needs to manually select which category the paper belong to. This webapp allow user to enter their abstract (i.e., a few lines of the summary of their important paper), then the code will analyze it and suggest a category it belongs (in astronomy). The classification task is done by a small neural network (trained by submitted abstracts).

Check the title for the webapp.

The code can be found here .

For the deployment detail (e.g., ow to deploy the webapp (with numpy and tensorflow) onto AWS Lambda ), I have written a blog post about it.

Demo of webapp (20 news group dataset)

webapp

(It may take a few seconds to load.)

Before fully implementing the above project, I tried to make a working demo. This demo takes two numbers (features) from users, and predict which type of Iris it should belong. While it is a textbook machine learning, I have made an web app.

Features:

• Data (model parameters) are hosted on Amazon S3
• Webapp is running on AWS Lambda using flask and scikit-learn.

The codes and reference are hosted on GitHub .