Made custom word embeddings from a series of essays I wrote analyzing the rhetoric of Houseki no Kuni (Land of the Lustrous) by Ichikawa Haruko, my favorite manga!
This project is part of an assignment for the natural language processing class I took in university on exploring pre-trained and custom word embeddings. I wanted to learn about the matrices produced by SVD that create the setting for capturing potential relationships between words from a corpus' vocabulary as represented by vectors.
I use TFIDF and SVD to create custom word embeddings based on a corpus of my manga analysis essays. Vector distances for select custom embeddings are visualized using PCA, and are also compared to graphs of vector distances between pre-trained embeddings of the same words but taken from GloVE's wiki-gigaword-100 dataset.