This work investigates a novel computational model of preverbal infant word learning in an attempt to create a more robust speech recognition system. Currently, the state-of-the-art can be extremely accurate when used in its optimal environment. However, when taken out of its comfort zone accuracy significantly deteriorates and does not come anywhere near human speech processing abilities, even for the simplest of tasks. We take inspiration from the ease with which newborns are able to learn words, with no apparent difficulty, and develop into expert communicators of their native language.