nerdimite / apprentice

An EdTech Solution for Educators and Teachers (Udacity Capstone Project)

Apprentice

Udacity Machine Learning Engineer Nanodegree Capstone Project

Apprentice is an automatic test assessment engine made as part of the Capstone Project in Udacity's Machine Learning Engineer Nanodegree.

The Project Report can be found here

Setup

Libraries Required

• Tensorflow 2.0 (Preferably GPU version)
• Tensorflow-Hub
• Scikit Learn
• Numpy
• Pandas
• tqdm
• Matplotlib
• Anvil-Uplink (pip install anvil-uplink)

Usage

Inference

The pretrained model can be used to perform inference and make predictions with the web app.

1. Run all the cells in Inference.ipynb
2. Open the Web Application at https://apprentice.anvil.app/. The input text boxes in the web app are prefilled to give an example.

Train

If you wish to re-train the model, you can refer Train.ipynb notebook.

References

• Dataset: https://www.quora.com/q/quoradata/First-Quora-Dataset-Release-Question-Pairs
• Benchmark Models: https://towardsdatascience.com/finding-similar-quora-questions-with-bow-tfidf-and-random-forest-c54ad88d1370