Louisa Best
Phonetic Spelling Correction Using Vector Space Models and Dimensionality Reduction
“Phonetic spelling correction algorithms such as Soundex and Metaphone offer simple phonetic matches but
do not handle the complexity of misspellings by individuals with intellectual disabilities. Jerome et al. (2010)
emphasised the need for specialised tools, while pronunciation models (Toutanova & Moore 2002) do not
fully address phonetic variability. Building on Wang et al. (2019), this project expands vector space models
to address phonetic spelling variability using dimensionality reduction.
Research Question/Objectives
How can vector space models and dimensionality reduction improve predictive spelling systems for individuals
with intellectual disabilities who type phonetically?
Objectives:
1. Develop a phonetic vector space model.
2. Apply dimensionality reduction (SVD, PCA) to find patterns.
3. Use Cosine Similarity for word prediction.
4. Evaluate the accuracy against existing algorithms.”
Louisa Best
Deakin University
Louisa Best is a second-year Computer Science student at Deakin University, specialising in Data Science. With a keen interest in computational intelligence and natural language processing (NLP), Louisa is passionate about all things computer science and mathematics.
Her current research addresses the communication barriers faced by individuals with intellectual disabilities, specifically through the development of phonetic spelling correction systems. This project employs vector space models, dimensionality reduction, and mathematical techniques to enhance accessibility and communication.
Louisa’s passion extends to analysing data, statistics, and applying mathematical approaches to solve complex problems. Louisa is dedicated to using her skills to contribute meaningful solutions to the field and is enthusiastic about future opportunities to make a significant impact.
