A Model of Semantics and Corrections in Language Learning

We present a computational model of language learning via a sequence of interactions be-tween a teacher and a learner. The utterances of the teacher and learner refer to shared sit-uations, and the learner uses cross-situational correspondences to learn to comprehend theteacher’s utterances and produce appropriate utterances of its own. We show that in this modelthe teacher and learner come to be able to understand each other’s meanings. Moreover, theteacher is able to produce meaning-preserving corrections of the learner’s utterances, and thelearner is able to detect them.We test our model with limited sublanguages of several natural languages in a commondomain of situations. The results show that learning to a high level of performance occurs aftera reasonable number of interactions. Moreover, even if the learner does not treat correctionsspecially, in several cases a high level of performance is achieved significantly sooner by a learnerinteracting with a correcting teacher than by a learner interacting with a non-correcting teacher.Demonstrating the benefit of semantics to the learner, we compare the number of interactionsto reach a high level of performance in our system with the number of similarly generatedutterances (with no semantics) required by the ALERGIA algorithm to achieve the same levelof performance.We also define and analyze a simplified model of a probabilistic process of collecting correc-tions to help understand the possibilities and limitations of corrections in our setting.