A cross-situational algorithm for learning a lexicon using Neural modeling fields

Abstract

Cross-situational learning is based on the idea that a learner can determine the meaning of a word by finding something in common across all observed uses of that word. Although cross-situational learning is usually modeled through stochastic guessing games in which the input data vary erratically with time (or rounds of the game), here we investigate the possibility of applying the deterministic Neural Modeling Fields (NMF) categorization mechanism to infer the correct object-word mapping. Two different representations of the input data were considered. The first is termed object-word representation because it takes as inputs all possible objectword pairs and weighs them by their frequencies of occurrence in the stochastic guessing game. A re-interpretation of the problem within the perspective of learning with noise indicates that the cross-situational scenario produces a too low signal-tonoise ratio, explaining thus the failure of NMF to infer the correct object-word mapping. The second representation, termed context-word, takes as inputs all the objects that are in the pupil's visual field (context) when a word is uttered by the teacher. In this case we show that use of two levels of hierarchy of NMF allows the inference of the correct object-word mapping.

Cognitively Inspired Neural Network for Recognition of Situations

“…Therefore, usually it is not obvious which label-word refers to which situation. This case of partial supervision is sometimes modeled as cross-situational learning (Fontanari et al, 2009). In future we will address similar learning of situations in parallel with proper associations among word-labels and situations.…”

Evaluation of Word Representations in Grounding Natural Language Instructions Through Computational Human-Robot Interaction

“…a word, through perceptual information [18]. Previous studies that investigated the use of crosssituational learning for grounding of objects [13,40] as well as spatial concepts [2,10,41] ensured that one word appears several times together with the same perceptual feature vector so that a corresponding mapping can be created [14]. However, natural language is ambiguous due to homonymy, i.e.…”

Action learning and grounding in simulated human–robot interactions

“…To ground manipulation actions in an unsupervised manner, i.e. without the need for a tutor, CSL (Section 2.2) can be used, which assumes that one word appears several times together with the same perceptual feature vector so that a corresponding mapping can be created (Siskind, 1996;Fontanari et al, 2009b;Smith et al, 2011). Previous studies investigated the use of CSL for grounding of objects and actions (Fontanari et al, 2009a;Taniguchi et al, 2017) as well as spatial concepts (Tellex et al, 2011;Dawson et al, 2013;Aly et al, 2017).…”