Computational models of tutor feedback in language acquisition

Abstract

This paper investigates the role of tutor feedback in language learning using computational models. We compare two dominant paradigms in language learning: interactive learning and cross-situational learning -which differ primarily in the role of social feedback such as gaze or pointing. We analyze the relationship between these two paradigms and propose a new mixed paradigm that combines the two paradigms and allows to test algorithms in experiments that combine no feedback and social feedback. To deal with mixed feedback experiments, we develop new algorithms and show how they perform with respect to traditional knn and prototype approaches.

Combining Unsupervised and Supervised Learning for Sample Efficient Continuous Language Grounding

“…Combining unsupervised and supervised grounding approaches has so far not received much attention despite the potential to combine their strengths and eliminate or at least reduce the impact of their shortcomings. Nevens and Spranger (2017) investigated the combination of cross-situational and interactive learning and came to the conclusion that the more feedback is provided, the faster new mappings are obtained and the higher the accuracy of the obtained mappings. While these findings, i.e., that feedback improves the accuracy and sampleefficiency, seem reasonable and intuitive, the employed crosssituational learning algorithm was very limited, thus, it is not clear whether feedback would have provided the same benefit, if a more sophisticated unsupervised grounding mechanism would have been employed.…”

“…Since in both studies one of the employed mechanisms, i.e. the unsupervised mechanism in (Nevens and Spranger, 2017) and the supervised mechanism in (Roesler, 2020a), were quiet limited, this study combines two mechanisms that have previously been shown to achieve state-of-the-art grounding results and evaluates whether their combination leads to better sample-efficiency and accuracy, while ensuring that supervision can be provided in a simple and natural way, and is not required to learn the correct groundings.…”

“…Examples are cross-situational learning (Siskind, 1996;Smith et al, 2011) based approaches that have been used to ground objects, actions, and spatial concepts (Dawson et al, 2013;Aly et al, 2017). Only limited work, i.e., (Belpaeme and Morse, 2012;Nevens and Spranger, 2017;Roesler, 2020a), has been done to compare or combine both approaches (see Section 2.3). Therefore, this study proposes and evaluates a hybrid grounding framework that combines both paradigms.…”

The development of a robust symbolic communication system for robots via embodied iterated imitation

“…They claimed that the agents can agree on a shared lexicon by a series of language games by acquiring rule-based inferences based on the results of the games. The models for tutor feedback mechanisms further discussed by Nevens and Spranger (2017). They analyzed two different feedback mechanisms, namely, gaze or pointing, and proposed a mixed paradigm that combines these two methods.…”