Robot acquisition of lexical meaning - moving towards the two-word stage

Abstract

We report on experiments and analyses dealing with the acquisition of lexical meaning in which prosodic analysis and extraction of salient words are associated with a robots sensorimotor perceptions in an attempt to ground these words in the robots own embodied sensorimotor experience. We focus here on three key areas, the selection of salient words based on prosodic clues, expression of words by the robot at a two-word stage to reflect learning and grammatically correct presentation, and an in-depth analysis of the relationship between words and the robots sensorimotor perceptions.

Robots That Say “No” Affective Symbol Grounding and the Case of Intent Interpretations

“…human language development as well as laying the groundwork for a new generation of the symbol grounding systems that go beyond the grounding of words and phrases with external physical referents. We reimplemented a language acquisition system very similar to another such system which had been successfully employed for the acquisition of nouns, adjectives, and two-word utterances [19]- [21]. This type of acquisition system relies on a tight temporal coupling between social mechanisms such as the establishment of joint attention on one hand, and some form of embodied machine learning on the other.…”

“…They were further asked to talk to the robot as if it were a prelinguistic child and told that it would express preferences for certain objects. The participants' instructions were largely identical to an earlier experiment on noun acquisition executed in the same premises and with the same robot which we will henceforth refer to as Saunders et al's [21] experiment. This previous work did not involve affective displays on part of the robot and focused on symbol grounding via sensorimotor association.…”

Grounded lexicon acquisition — Case studies in spatial language

“…Also in the ICDL/Epirob community this is a recurring theme that has received considerable attention (e.g. see [9], [10] for recent examples). However, many of these studies are done in pure simulation, or presuppose shared meaning spaces and/or shared world models.…”

Robots Learning to Say “No”

“…We will therefore frequently refer to the latter as rejection experiment and will use it as comparative "benchmark". Both negation experiments crucially involve affect or motivation, and both build up on "non-affective" work on robotic language acquisition conducted by Saunders et al [38][39][40]. The experimental setup of both negation experiments and the work of Saunders et al are similar to such a degree that they allow us to conduct comparisons between participants' speech between these experiments.…”

“…We used identical instructions in both the rejection [14] as well as the present prohibition experiment both of which, in turn, were very similar to Saunders' experiment [38]. The majority of instructions in both negation experiments as well as Saunders' experiment [38] were identical: participants were told that they ought to teach the robot Deechee about the objects on the table. We tried to prime participants into adopting a style of speech akin to child-directed speech (CDS, [15,27,43]) by telling them to imagine Deechee to be a two-year-old.…”