Author

David Duffrin

Date of Award

2017

Document Type

Thesis

Degree Name

Bachelors

Department

Natural Sciences

First Advisor

Gillman, David

Area of Concentration

Computer Science

Abstract

An advanced learner of the language or native speaker can extrapolate the pronunciations for a given kanji to estimate the reading of a never before seen word. Most textbooks discourage learners of Japanese from finding patterns and instead advise memorization of a kanji in each context to gain proficiency in learning. I disagree with this approach and designed a Naive Bayes Classifier that is trained on a simple Japanese dictionary. The model is able to impute the pronunciation of Japanese kanji with 90% accuracy in limited testing. The probabilities from the Naive Bayes Classifier can be easily read and understood by a Japanese learner to quickly gain proficiency in the language. As the end product, my model will help learners gain the intuition needed to estimate the pronunciation of a given kanji in an unfamiliar context.

Share

COinS