Improving open text matching in a communication serious game

Abstract

I analyze and implement state-of-the-art natural language processing models for open text understanding to improve the matching of open text input in a serious game that uses custom scenarios created for training communication skills, called Communicate. Previous work in matching open text input in this serious game used a scenario specic corpus, a corpus containing all the words used in the particular scenario, to match open text input to a scripted statement. This scenario-specic corpus contains mathematical representations for each word appearing in the scenario. The goal of this thesis is to expand on this previous work by exploring state-of-the-art word embeddings and implementing relevant models that use scenario specic information to try to improve the open text matching process.