Continuous measure of word learning supports associative model

Abstract

Abstract-Cross-situational learning, the ability to learn word meanings across multiple scenes consisting of multiple words and referents, is thought to be an important tool for language acquisition. The ability has been studied in infants, children, and adults, and yet there is much debate about the basic storage and retrieval mechanisms that operate during cross-situational word learning. It has been difficult to uncover the learning mechanics in part because the standard experimental paradigm, which presents a few words and objects on each of a series of training trials, measures learning only at the end of training after several occurrences of each word-object pair. Thus, the exact learning moment-and its current and historical contextcannot be investigated directly. This paper offers a version of the cross-situational learning task in which a response is made each time a word is heard, as well as in a final test. We compare this to the typical cross-situational learning task, and examine how well the response distributions match two recent computational models of word learning.

Cross-situational noun and adjective learning in an interactive scenario

“…The proposed approach, while not aiming at modelling human behaviour, is compatible with the hypothesis that humans use associative methods for cross-situational learning [9], [11]. Like these models, it uses statistics on word occurrences and their associations with referents.…”

“…The other approach is based on associative learning and memorizes an association table between all words and referents that is updated after each observation [9]. Empirical evaluation supports the idea that associative models better match human behaviour [9], [11]. An approach based on a Bayesian framework has been proposed by [12] which is moreover able to integrate social cues in learning and was shown to outperform a simple associative model on a dataset of parent-children interactions.…”

An experimental comparison between NMF and LDA for active cross-situational object-word learning

“…Several models of cross-situational learning have been proposed using different techniques such as hypothesis testing and associative learning [7], Expectation-Maximisation [6] or measures of co-occurences and mutual information [5]. In this paper, we focus on latent topic discovery approaches [8] that are well suited to this problem.…”

“…Several models of cross-situational learning have been proposed (e.g., [5], [6], [7]). In particular, this problem can be "Here are a red ring, a yellow cup and a green lego."…”

“…We focus on their performance evolution while incrementally learning from a limited set of examples and explore whether they are well suited to define a value function for active learning. Unlike most models of crosssituational learning [4], [5], [6], [7], [12], [15], we moreover focus on the learning of object's descriptive features instead of only its identity and deal with the linguistic ambiguity besides the classical referential ambiguity in experimental settings. Tackling both these ambiguities, our model lays the basis for interactive learning of concept for developmental robots.…”