- Research Program Mentor
PhD Doctor of Philosophy candidate
natural language processing, machine learning, broadly engineering, audio processing, computational linguistics
Could language models experience synesthesia?
Synesthesia is the experience in which some people (synesthetes) experience certain unrelated sensory or cognitive stimulation when experiencing another one. (For example, seeing colors when listening to music.) Word embedding models such as Word2Vec and GloVe are trained over large corpora of language to assign similar high-dimensional value vectors (word embeddings) to similar words. (For example, "apple" and "peach" would have more similar word embeddings than would "apple" and "swimming".) The goal of this project would be to determine whether these trained word embedding models might possess common synesthetic associations such as the letter A to the color red. How might these association, if any, vary across word embedding models trained on different languages?