EduSelect AI

EduSelect AI is a web-based tool dedicated to helping students at Cal Poly (California Polytechnic State University) discover their ideal college major based on their course and program preferences. This project leverages state-of-the-art natural language processing (NLP) techniques, including BERT-based word embeddings and various classification models, to provide tailored major recommendations.

Features

• User-Friendly Web Application: Access the tool through a user-friendly web application that simplifies the major exploration process.
• Input Course and Program Descriptions: Users input course descriptions and program descriptions, helping the system understand their academic interests.
• Contextual Embeddings: Utilizes BERT (Bidirectional Encoder Representations from Transformers) to generate contextual embeddings for user descriptions, capturing the nuanced meaning and context of the input data.
• Multi-Classifier System: Employs a multi-classification system with three distinct classifiers:
  • K-Nearest Neighbors (KNN): Recommends majors based on the similarity of user descriptions to existing course and program data.
  • Multi-Layer Perceptron (MLP): Utilizes a neural network to map embedded information to potential college majors.
  • DistilBERT Classifier: Leverages a fine-tuned DistilBERT model to predict the most suitable majors.

Results

The project has achieved promising results in major recommendation accuracy, as demonstrated by the following performance metrics: - BERT + KNN: Achieved an accuracy of 0.47 in major recommendations. - BERT + Major MLP: Improves accuracy to 0.71, providing more refined major suggestions. - DistilBERT Classifier: Delivers the highest accuracy of 0.74, offering precise major recommendations.

Getting Started

1. Set up python environment:

conda env create --file environment.yml
conda activate csc-570

Project Layout

• embeddings: Sklearn-style transformers that encode natural language into latent embedding vectors.
• classifers: Model architectures for classifying college majors.
• test.py: Evaluation and demo code for all models.
• train.py: Training loops for models.
• helper.py: Utility methods for loading and preprocessing data

Command Line Demo

Run the demo() function in test.py

Web Server

The web server provides a RESTful API for getting major recommendations:

python api.py
curl --location --request POST 'http://127.0.0.1:5000/recommend' \
--header 'Content-Type: application/json' \
--data-raw '{
    "data": "Construction is cool!"
}'

...

[
    "Construction Management",
    "Architectural Engineering",
    "City and Regional Planning"
]
``