Overview
As part of our coursework in CPEN 221, we were put in teams to complete several mini-projects throughout the term. Our first project involved creating text prediction models as well as RateMyProfessor review analysis, and I led the completion of these more challenging tasks. In Java, we created new classes, programmed functions, added specifications and developed a comprehensive test suite using Gradle for each milestone. Code on Github here: https://github.com/Eric-Ptn/cpen221-miniproject1
Design
- Created a computer model to predict a word's synonyms using a cosine similarity measure between strings, trained on 20+ texts from classic literature
- Led the development of a RateMyProfessor sentiment analyzer, through a Bayesian probability calculation that maps word frequency to likely ratings
- Helped create an algorithm to generate potential search terms given a query, sorted by their weights
Results
- Synonym prediction model correctly guesses over 80% of synonyms, even with a limited dataset
- 85% accuracy in predicting the RateMyProfessor score of test review ratings
