A Model of Semantics and Corrections in Language Learning

We present a computational model of language learning via a sequence of interactions between a teacher and a learner. The utterances of the teacher and learner refer to shared situations, and the learner uses cross-situational correspondences to learn to comprehend the teacher’s utterances and produce appropriate utterances of its own. We show that in this model the teacher and learner come to be able to understand each other’s meanings. Moreover, the teacher is able to produce meaning-preserving corrections of the learner’s utterances, and the learner is able to detect them. We test our model with limited sublanguages of several natural languages in a common domain of situations. The results show that learning to a high level of performance occurs after a reasonable number of interactions. Moreover, even if the learner does not treat corrections specially, in several cases a high level of performance is achieved significantly sooner by a learner interacting 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 interactions to reach a high level of performance in our system with the number of similarly generated utterances (with no semantics) required by the ALERGIA algorithm to achieve the same level of performance. We also define and analyze a simplified model of a probabilistic process of collecting corrections to help understand the possibilities and limitations of corrections in our setting.

[1]  Dana Angluin,et al.  Inference of Reversible Languages , 1982, JACM.

[2]  Pierre Dupont,et al.  Stochastic Grammatical Inference with Multinomial Tests , 2002, ICGI.

[3]  A. M. Schaerlaekens,et al.  The two-word sentence in child language development : a study based on evidence provided by Dutch-speaking triplets , 1973 .

[4]  Isabelle Tellier,et al.  Meaning Helps Learning Syntax , 1998, ICGI.

[5]  Jeffrey Mark Siskind,et al.  Lexical Acquisition in the Presence of Noise and Homonymy , 1994, AAAI.

[6]  José Oncina,et al.  Learning Stochastic Regular Grammars by Means of a State Merging Method , 1994, ICGI.

[7]  G. Marcus.  Negative evidence in language acquisition , 1993, Cognition.

[8]  Makoto Kanazawa.  Learnable Classes of Categorial Grammars , 1998 .

[9]  Brian Scassellati,et al.  Using sentence context and implicit contrast to learn sensor-grounded meanings for relational and deictic words: the twig system , 2008 .

[10]  E. Mark Gold,et al.  Language Identification in the Limit , 1967, Inf. Control..

[11]  E. Clark,et al.  Adult reformulations of child errors as negative evidence , 2003, Journal of Child Language.

[12]  Eve V. Clark,et al.  The Lexicon in Acquisition , 1996 .

[13]  R. Treiman,et al.  Brown & Hanlon revisited: mothers' sensitivity to ungrammatical forms , 1984, Journal of Child Language.

[14]  Eve V. Clark,et al.  The principle of contrast: A constraint on language acquisition. , 1987 .

[15]  Brian Scassellati,et al.  Robotic vocabulary building using extension inference and implicit contrast , 2009, Artificial Intelligence.

[16]  Colin de la Higuera,et al.  Probabilistic DFA Inference using Kullback-Leibler Divergence and Minimality , 2000, ICML.

[17]  Alex Pentland,et al.  Learning words from sights and sounds: a computational model , 2002, Cogn. Sci..

[18]  Catherine L. Harris,et al.  The human semantic potential: Spatial language and constrained connectionism , 1997 .

[19]  Leonor Becerra-Bonache,et al.  Learning Meaning Before Syntax , 2008, ICGI.

[20]  L L Travis,et al.  Limits on negative information in language input , 1989, Journal of Child Language.

[21]  Nick Chater,et al.  Distributional Information: A Powerful Cue for Acquiring Syntactic Categories , 1998, Cogn. Sci..

[22]  Andreas Stolcke,et al.  Miniature Language Acquisition: A Touchstone for Cognitive Science , 2002 .

[23]  Alexander Clark,et al.  Unsupervised Language Acquisition: Theory and Practice , 2002, ArXiv.

[24]  Jerome A. Feldman,et al.  When push comes to shove: a computational model of the role of motor control in the acquisition of action verbs , 1997 .

[25]  C. Snow,et al.  Feedback to first language learners: the role of repetitions and clarification questions , 1986, Journal of Child Language.

[26]  Dana H. Ballard,et al.  A multimodal learning interface for grounding spoken language in sensory perceptions , 2004, ACM Trans. Appl. Percept..

[27]  Francisco Casacuberta,et al.  Probabilistic finite-state machines - part II , 2005, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[28]  Jane C. Hill.  A computational model of language acquisition in the two-years-old , 1982 .

[29]  Dana Ron,et al.  On the learnability and usage of acyclic probabilistic finite automata , 1995, COLT '95.

[30]  Yasubumi Sakakibara,et al.  Efficient Learning of Context-Free Grammars from Positive Structural Examples , 1992, Inf. Comput..

[31]  Terry Regier,et al.  The Emergence of Words: Attentional Learning in Form and Meaning , 2005, Cogn. Sci..

[32]  Brian Scassellati,et al.  A Robot That Uses Existing Vocabulary to Infer Non-Visual Word Meanings from Observation , 2007, AAAI.

[33]  J. Siskind.  A computational study of cross-situational techniques for learning word-to-meaning mappings , 1996, Cognition.