IEEE ICDL-EPIROB Workshop on Language Learning
This is the website of the 2016 workshop. The website of the 2017 edition of this workshop can be found here.
Dates and Place
Date: Monday, September 19th 2016
Time: 08:50am
Place: University of Cergy-Pontoise, Site “Saint-Martin” Building A, 5th floor (map)
Room: 546
This workshop is part of the 2016 IEEE ICDL-EPIROB conference, that takes place at Cergy-Pontoise, France (near Paris).
Schedule
08:50 - 09:00 Introduction
09:00 - 09:45 Contexts for word learning at scale. Linda Smith (Indiana University)
09:45 - 10:00 Language learning across tasks. Katharina J. Rohlfing (Paderborn University)
10:30 - 11:00 Coffee Break
11:00 - 11:45 Sensory-motor behaviors during child-parent toy play predict word learning. Chen Yu (Indiana University)
11:45 - 12:30 Language socialization in three radically different cultures. Paul Vogt (Tilburg University)
12:30 - 14:00 Lunch Break 14:00 - 14:45 Language acquisition by robots: How realistic should it be? Thierry Poibeau and Isabelle Tellier (CNRS)
14:45 - 15:30 Reverse engineering infant language acquisition. Emmanuel Dupoux (ENS, CNRS) 15:30 - 16:00 Coffee Break 16:00 - 16:15 Poster elevator pitch
16:15 - 17:00 Poster session
17:00 - 19:30 Informal lab tour and welcome drink
Topic
Children acquire language by interacting with their caregivers and others in their social environment. When children start to talk, their sensory-motor intelligence (visual perception, body movement, navigation, object manipulation, auditory perception and articulatory control) is already reaching a high level of competence. These competences and growing representations provide a basis for the ongoing development of communication. Importantly, communication is based on representations and skills that have started to develop much earlier and that are shaped already by the first social interactions. These interactions are multimodal in nature and vary across contexts. The contexts vary not only across developmental time and situations within individuals, but also between individuals, socio-economic groups and cultures. Continuously, representations become further enriched in ongoing interactions and across different contexts.
Even though there are various efforts in developmental robotics to model communication, the emergence of symbolic communication is still an unsolved problem. We are still lacking convincing theories and implementations that show how cooperation and interaction knowledge could emerge in long-term experiments with populations of robotic agents.
Importantly, continuously acquiring knowledge in different contexts and being able to further enrich the underlying representations provides a potential powerful mechanism (cross-situational learning) which is already well recognized in learning in children. Still, we need to know more about how children recognize contexts and how their language learning benefits from different language use varying across contexts.
Objectives
Addressing the emergence of communication requires combining and integrating knowledge from diverse disciplines: developmental psychology, robotics, artificial language evolution, complex systems science, computational linguistics and machine learning. The goal of the workshop is to bring together researchers from these areas in order to discuss current findings from experimental studies and to transfer hypothetical insights into potential mechanism for modelling approaches.
In particular, the workshop has the following three objectives:
Set up a roadmap for multidisciplinary research exploiting multimodal language learning across contexts. Our discussions will focus on the notion of “context” in order to specify some parameters that broaden our understanding of how these may be constructed (or emerge) or how these may be utilized in, for instance, cross-situational learning.
The ability to cooperate as well as to communicate is assumed to rely on rich embodied representations. One of the key objectives of the workshop will be to understand possible joint representations (e.g., sensorimotor schemas and constructions). In particular the workshop will focus on the overlap in these representations and on the question how such representations can serve different tasks as in motor control or when recruited in communication. Secondly, how the representations are interacting in different stages of development?
The workshop will examine cognitive architectures for learning and acquisition strategies with a special emphasis on architectures that allow modeling the interplay of different strategies on all levels of development e.g. the acquisition and evolution of interaction patterns, shaping and change of word meanings, schematisation of constructional knowledge etc. How can we build learning mechanisms that embrace the complexity of information and its variability across contexts?
Invited Speakers
Katharina J. Rohlfing (Paderborn University, Germany): Language learning across tasks. In this talk, I will present results from a slow mapping study, in which children were trained to apply a new word on scenes from either pictures or toys. The results reveal the complexity of factors that influence generalization of word categories: children’s learning performance during an experiment depends on the task demands and on the stimuli used. A given performance does not necessarily reflect a stable representation; it may rather be a reflection of the child’s ability to handle the task demands. It seems that very specific training (i.e., with particular items and within a particular task) supports children’s slow mapping but interfaces with the children’s experience to process linguistic and task demands. In this sense, the development of the semantic network does not seem to be an exclusive product of overall exposure to language. Instead, the initial representation also binds memories about the pragmatic circumstances in which the word occurs.
Linda B. Smith (Indiana University, USA): Contexts for word learning at scale. The average toddler in the United States hears 20,000 words a day. That toddler is also awake about 12 hours a day viewing and interacting with the objects in that world. In brief, the language and visual data that feeds language learning is massive. That data, for both words and objects, has a distributional structure that is not uniform, lumpy, and bursty. This talk will consider how diversity and consistency in the input data for word learning supports early vocabulary development and how the properties of that distributional structure of words and visual objects in the learning environment may be responsible for individual differences in early vocabulary development.
Paul Vogt (Tilburg University, the Netherlands): Language socialization in three radically different cultures. This talk presents findings from a longitudinal observational study on the language socialization of infants in the Netherlands, urban Mozambique and rural Mozambique. This study crucially shows how language socialization of infants in non-Western cultures is radically different from what is observed and known in Western cultures, which require us to carefully reconsider our theoretical models of language acquisition.
Chen Yu (Indiana University, USA): Sensory-motor behaviors during child-parent toy play predict word learning. Children learn words with a social partner. This talk focuses on the quality of parent-child interactions in the belief that it is the foundation of early word learning. I will present recent findings to support the argument that parents and young children dynamically couple behavior and attention when engaging with objects, and these moment-to-moment behaviors are fundamental building blocks for smooth social coordination which predict language development.
Thierry Poibeau and Isabelle Tellier (CNRS, Paris, France). Language acquisition by robots: How realistic should it be? In this talk, we will compare the language acquisition process by humans with recent artificial models implemented in robotic environments. We will first describe things we know (and also, if at all possible, things we don’t know) about language acquisition. We will then examine recent models and proposals aiming at formalizing the acquisition process in robotic environments. We will discuss to what extent these models fit with actual observations in natural environments and we will then conclude by asking whether a realistic model is possible or even desirable.
Emmanuel Dupoux (CNRS/ENS, Paris, France). Reverse engineering infant language acquisition. During their first years of life, infants learn the language(s) of their environment at an amazing speed despite large cross cultural variations in amount and complexity of the available language input. Here, we analyze the conditions under which recent progresses in Machine Learning can contribute to our scientific understanding of early language development. We argue that instead of defining a sub-problem or simplifying the data, computational models should address the full complexity of the learning situation, and take as input the raw sensory signals available to infants. This implies that (1) accessible but privacy-preserving repositories of home data be setup and widely shared, and (2) models be evaluated at different linguistic levels through a benchmark of psycholinguist tests that can be passed by machines and humans alike, (3) linguistically and psychologically plausible learning architectures be scaled up to real data using probabilistic/optimization principles from machine learning. We discuss the feasibility of this approach and present preliminary results.
Format
The workshop is planned is a full day workshop with three sessions of three talks each (45 minutes incl discussion each). We invited a total of 6 senior researchers.
The workshop finishes with a poster session (at the end and during the first coffee break). We encourage the submission of poster abstracts for the workshop. We want to give young people a chance to present their (ongoing) work. But we also want to provide a forum for relevant work that has recently been published in journals and other conferences. Abstracts will be reviewed by the organizers. Suitable posters will be invited to submit their work to an upcoming special issue of Frontiers in Neurorobotics.
Call for Papers
We invite the submission of abstracts (anywhere between 200 words to 2 pages) related to all aspects of language learning. Accepted abstracts are presented in a poster session. We particularly encourage researchers to submit abstracts of ongoing work, as well as recently published work related to all aspects of language learning in artificial systems and humans.
Abstracts are reviewed for suitability by the organizers. If in doubt about whether your poster is in scope for the workshop, please contact us directly.
Publication: Suitable posters will be invited to submit their work to an upcoming special issue of Frontiers in Neurorobotics.
Submission: two page abstracts
Deadline: August 19th, 2016
Please send you abstracts to languagelearningcontact@gmail.com
Organizers
Chen Yu (Indiana University, USA)
Katharina J. Rohlfing (Paderborn University, Germany)
Malte Schilling (CITEC Bielefeld, Germany)
Michael Spranger (Sony Computer Science Laboratories Inc, Japan)
Paul Vogt (Tilburg University, the Netherlands)
Contact
If you have any questions, comments or feedback, please contact Michael Spranger at michael [dot] spranger [at] gmail [dot] com