Trieste Encounters on Cognitive Science, Language Learning

Jul 7, 2016, 8:00 AM → Jul 15, 2016, 6:00 PM Europe/Rome

Meeting room (7th floor) (SISSA main building)

Alessandro Treves (SISSA), Davide Crepaldi (SISSA), Luca Piretti (SISSA)

The Trieste Encounters on Cognitive Science (TEX) are a yearly summer initiative by SISSA, aiming at bringing together scientists from all over the world at all levels of their careers to enjoy science, sunshine, food&wine and, ultimately, life. 

This year's edition will focus on language learning construed broadly enough to include the progressive mastering of: (i) a mother tongue as a child; (ii) a second or artificial language as an adult; and (iii) specific cognitive mechanisms for facing language input more efficiently (e.g., chunking), as in artificial networks. We strive to put together different perspectives -- neuroscience, linguistic theory, cognitive psychology and computational modelling -- so as to promote the cross–breeding that language science desperately needs. We are also trying to open up our perspectives and look for ideas/concepts/data coming from outside language, which still may point to general learning mechanisms that may also apply to some linguistic phenomena.

The TEX2016 philosophy is focusing on open issues, rather than celebrating achievements. Despite the progress that we have made as a field in the last 50 years, we are clearly still far away from any mechanistic understanding of how our brain/mind expresses such a wonderful capacity as language. We want to recognize our failures in the past, look straight to the most important unanswered questions out there, and join our efforts to find new ways of tackling them. For this, we need both a solid recap of the knowledge we have accumulated over the years, and a look forward to the most promising new approaches and theories.     

In line with this philosophy, TEX2016 wants to be a combination between a summer school (i.e., we will cover the fundamentals so as to address students approaching the field) and a symposium (i.e., we will explore the frontiers of knowledge and discuss open issues in the field). Accordingly, we have invited scholars to serve as teachers (see below) -- they will offer classes on the established concepts and body of evidence in their own specialty; but we also issue an open call for research talks, so that participants will be exposed to cutting-edge pieces of research, and hopefully cut some edge themselves. We count on a diversity of participants of varying seniority and background to self-organize and find the best way to interact and maximize their contribution to - and benefit from - the event. You can apply as a student or register as a participant by clicking on Registration on the left; and you can contribute an abstract by clicking on Call for Abstracts, again on the left.  

The details of the program will be defined based on the response to the open call; but the idea is to have classes on fundamentals in the mornings and research talks in the afternoons, with plenty of time remaining for social (or asocial) activities.

TEX2016 will take place between Thursday July 7 and Friday July 15, 2016. It will be held at SISSA main campus, via Bonomea 265, Trieste. Information on how to reach us can be found here.

TEX2016 will feature Alex Cristia (Paris), Colin Davis (Bristol), Edmund Rolls (Warwick), Gareth Gaskell (York), Kathy Rastle (Royal Holloway, London), Max Louwerse (Tilburg), Na'ama Friedmann (Tel Aviv), Pino Longobardi (York), Scott Johnson (UCLA), and Vito Pirrelli (Pisa). Tim Shallice, the founding father of cognitive science at SISSA (and perhaps more widely), will also join us as a discussant. Yann LeCun (New York Univerisity and chief of AI research at Facebook) will give us the closing lecture.

We look forward to meeting you in Trieste!

Davide Crepaldi, Alessandro Treves, and the TEX2016 team

Thursday, July 7

Welcome

Conveners: Prof. Alessandro Treves (SISSA), Prof. Davide Crepaldi (SISSA)

1 Longobardi I

Speaker: Prof. Pino Longobardi (University of York)

Slides longobardi_tex2016.pdf

12:30 PM Lunch

2 Johnson I

Speaker: Prof. Scott Johnson (UCLA)

Slides johnson_tex2016_1.pdf

4:00 PM Coffee break

3 Longobardi II

Speaker: Prof. Pino Longobardi (University of York)

Slides longobardi_tex2016.pdf

Friday, July 8

4 Gaskell I

Speaker: Prof. Gareth Gaskell (University of York)

Slides gaskell_tex2016.txt

10:30 AM Coffee break

5 Johnson II

Speaker: Prof. Scott Johnson (UCLA)

Slides johnson_tex2016_2.pdf

12:30 PM Lunch

Contributed talks I

3:30 PM Coffee break

Contributed talks II

Saturday, July 9

6 Friedmann I

Speaker: Prof. Na'ama Friedmann (Tel Aviv University)

10:30 AM Coffee break

7 Gaskell II

Speaker: Prof. Gareth Gaskell (University of York)

Slides gaskell_tex2016.txt

12:30 PM Lunch

Contributed talks III

3:50 PM Coffee break

8 Information Theory and Language

Slides tex2016_Brasselet.pdf

Sunday, July 10

9 Friedmann II

Speaker: Prof. Na'ama Friedmann (Tel Aviv University)

10:30 AM Coffee break

10 Cristia I

Speaker: Dr Alex Cristia (CNRS and Ecole Normale Superieure)

Slides cristia_tex2016_1.pdf

summary Many long-standing questions in the study of child language and cognition remain elusive because of low signal-to-noise ratios and/or limited ecological validity of traditional empirical methods. Recent technological developments have augmented our toolkit with several measures that promise to enhance our empirical grasp. This talk seeks to inform the audience of the strengths and weaknesses of a selection novel methodologies (in terms of populations, topics, and designs). To be most informative, I center on a single topic: language acquisition in early childhood. This allows me to showcase three such methods, which stand to contribute unique insight into aspects ranging from a description of the spoken input and output, to the neural correlates of language processing. I introduce the method and illustrate its use with recent work from my own as well as other labs. The first method, LENA, consists of daylong recordings analyzed with largely unsupervised algorithms to provide a unique view of the child's verbal interactions with those around him/her. This is one of a series of emergent methods that try to capture children's behaviors in a naturalistic environment in all their richness, another example being that of head-mounted cameras. The second method involves electronic devices with a touch-screen. As children age, such an active and interactive method is more likely to uniquely engage their attention. Moreover, such a method could eventually allow precise language perception measurements outside the lab. Finally, fNIRS is an inexpensive neuroimaging technique, reputed to combine accurate localization of activations and a considerable resistance to movement artifacts. While the technique itself is far from new, its application to the description of brain networks involved in language perception and production, and cognition in general, remains relatively novel.

12:30 PM Lunch

Science in the Karst

Monday, July 11

11 Cristia II

Speaker: Dr Alex Cristia (CNRS and Ecole Normale Superieure)

Slides cristia_tex2016_2.pdf https://osf.io/r694c/

10:30 AM Coffee break

12 Rule Extraction and Generalisation in Language Learning

Much of language learning consists of discovering regularities across multiple exemplars, which can be used to interpret or produce new linguistic items. Morphology presents an important example of this discovery process. Though we are rarely exposed to prefixes and suffixes in isolation, our repeated exposure to them in different words (e.g. banker, runner, teacher) allows us to develop an understanding of their functionality, in such a way as to use them to create new words (e.g. tweeter; someone who tweets). In this talk, I will describe work that my group has conducted using artificial word learning paradigms, in which we have discovered some key constraints on this discovery process. I then relate the principles that have emerged in these laboratory studies to what is known about children's acquisition of morphological knowledge.

Speaker: Prof. Kathy Rastle (Royal Holloway University of London)

12:30 PM Lunch

A view from the other side

Tuesday, July 12

13 Mapping Visual Symbols onto Spoken Language Forms

Learning to read is arguably the most important challenge for the child starting school. Mapping new visual symbols onto existing spoken and conceptual forms also presents an interesting challenge for the brain. In this talk, I discuss some of the work that my laboratory has been conducting using artificial language learning approaches to understand how the brain solves this challenge. This work has allowed us to discover how the brain captures different forms of systematicity within the writing system, how the nature of the writing system impacts on emerging representations, and how the acquisition of literacy impacts on spoken language representations. I conclude by relating this work back to discoveries using more naturalistic methods, and argue that these two forms of evidence can be highly complementary in the study of reading acquisition.

Speaker: Prof. Kathy Rastle (Royal Holloway University of London)

10:30 AM Coffee break

14 Rolls I

Speaker: Prof. Edmund Rolls (University of Oxford)

Paper rolls_paper_tex2016.pdf

Slides rolls_tex2016.txt

12:30 PM Lunch

15 From memory processes to lexical self-organisation: a biologically-motivated integrative view of the morphological lexicon

According to “Words and Paradigm” approaches to morphological competence (Blevins 2006 among others), mastering the inflectional system of a language amounts to acquiring an increasing number of constraints on how paradigm are filled in by full word forms (see Ackerman et al. 2009; Finkel & Stump 2007; Pirrelli & Battista 2000; Matthews 1991; among others). Linguistic and developmental evidence on word paradigms has met recent developments in Computational Linguistics and Neurolinguistics. Self-organising artificial neural networks (Kohonen 2001; Pirrelli et al. 2015, Marzi & Pirrelli 2015) have offered an algorithmic account of the hypothesis that the mental lexicon is a highly-redundant, dynamic store of full words, which get co-activated and compete for selection during lexical processing. At the same time, recent advances in understanding the neuro-anatomical areas supporting memory (Wilson 2001; D’Esposito 2007; Ma et al. 2014) have showed that working memory consists in the transient activation of long-term memory structures, controlled and maintained by the integration of auditory-motor circuits in the perisylvian network (Catani et al. 2005; Shalom & Poeppel 2008). All these developments converge on the idea that stored lexical representations are in fact the long-term by-product of their processing history. In the talk, we illustrate simulative evidence supporting these insights and explore their theoretical implications for models of the mental lexicon. References Ackerman, Farrell; Blevins, James & Malouf, Robert 2009. Parts and wholes: implicative patterns in inflectional paradigms. In Blevins, James P. & Blevins, Juliette (eds.), Analogy in Grammar. Oxford: Oxford University Press. 54-82. Catani, Marco; Jones, Derek K. & ffytche, Dominic H. 2005. Perisylvian language networks of the human brain. Annals of Neurology 57. 8-16. D’Esposito, Mark 2007. From cognitive to neural models of working memory. Philosophical Transactions of the Royal Society B. Biological Sciences 362. 761-772. Finkel, Raphael & Stump, Gregory. 2007. Principal parts and morphological typology. Morphology 17. 39-75. Kohonen, Teuvo 2001. Self-organizing maps. Berlin Heidelberg: Springer-Verlag. Ma, Wei J.; Husain, Masud & Bays, Paul M. 2014. Changing concepts of working memory. Nature neuroscience 17(3). 347-356. Marzi, Claudia & Pirrelli, Vito 2015. A Neuro-Computational Approach to Understanding the Mental Lexicon. Journal of Cognitive Science 16 (4). 491-533. Pirrelli, Vito; Ferro, Marcello & Marzi, Claudia 2015. Computational complexity of abstractive morphology. In Baerman, Matthew; Brown, Dustan % Corbett, Greville (eds.). Understanding and Measuring Morphological Complexity. Oxford: Oxford University Press. 141-166. Pirrelli, Vito & Battista, Marco 2000. The paradigmatic dimension of stem allomorphy in Italian verb inflection. Italian Journal of Linguistics 12. 307-379. Matthews, Peter H. 1991. Morphology (second edition). Cambridge UK: Cambridge University Press. Shalom, Dorit B. & Poeppel, David 2008. Functional Anatomic Models of Language: Assembling the Pieces. The Neuroscientist 14. 119-127. Wilson, Margaret. 2001. The case of sensorimotor coding in working memory. Psychonomic Bulletin and Review 8(1). 44-57.

Speaker: Dr Vito Pirrelli (CNR, Pisa)

Slides pirrelli_tex2016_1.pdf

4:00 PM Coffee break

Contributed talks V

Wednesday, July 13

16 Word co-activation and competition and the perception of morphological structure

The cognitive literature on similarity-based principles of word association has greatly contributed to understanding effects of family size and frequency of neighbouring words on a variety of word processing tasks: non-word repetition (Vitevitch et al. 1997; Vitevitch & Luce 1998), recall from verbal short-term memory (Gathercole et al. 1997), phoneme identification (Pitt & McQueen 1998) and word recognition (Luce 1986; Luce & Pisoni 1998). Beyond specific differences depending on the nature of the input stimuli (e.g. acoustic vs. visual) and the processing requirements of the task (e.g. word recognition vs. word production), an interesting general pattern of reversal emerges: neighbours have facilitative effects on spoken word production and inhibitory effects in spoken word recognition. Furthermore, the frequency distribution of neighbours plays an important role in determining whether competition/co-activation effects are facilitative or inhibitory: high-frequency neighbours tend to exert an inhibitory effect on some processing tasks, while low-frequency neighbours facilitating execution of the same tasks. In the talk, we consider what competition-co-activation effects in lexical processing can tell us about the emergence of structure in morphologically complex words, based on evidence from highly inflected languages (Marzi et al. 2016) and English word compounding (Gagné & Spalding (in press); Ferro et al. 2016). References Ferro, Marcello; Marzi, Claudia; Pirrelli, Vito; Gagné, Christina & Spadling, Thomas 2016. One word or two? Discriminative effects of word entrenchment and competition on processing compounds and pseudo-compounds. 17 International Morphology Meeting, 18-21 February 2016, Vienna. Gagné, C.L. & Spalding, T.L. (in press). Effects of morphology and semantic transparency on typing latencies in English compound and pseudo-compound words. Journal of Experimental Psychology: Learning, Memory, and Cognition. Gathercole, Susan E.; Hitch, Graham J.; Service, Elisabet S. & Martin, Amanda J. 1997. Phonological short term memory and new word learning in children. Developmental Psychology 33. 966-979. Luce, Paul A. 1986. A computational analysis of uniqueness points in auditory word recognition. Perception and Psychophysics 39. 155–158. Luce, Paul A. & Pisoni, David B. 1998. Recognizing spoken words: The neighborhood activation model. Ear and hearing 19 (1). 1-36. Marzi, Claudia; Ferro, Marcello; Cardillo, Franco Alberto & Pirrelli Vito 2016. Effects of frequency and regularity in an integrative model of word storage and processing. In Marzi C. and Pirrelli V. (eds.), Italian Journal of Linguistics 28(1). Pitt, Mark A. & McQueen, James M. 1998. Is compensation for coarticulation mediated by the lexicon? Journal of Memory and Language 39(3). 347-370. Vitevitch, Michael S. & Luce, Paul A. 1998. When words compete: Levels of processing in spoken word recognition. Psychological Science 9. 325–329. Vitevitch, Michael S.; Luce, Paul A.; Charles-Luce, Jan & Kemmerer, David 1997. Phonotactics and syllable stress: Implications for the processing of spoken nonsense words. Language and Speech 40. 47–62.

Speaker: Dr Vito Pirrelli (CNR, Pisa)

Slides pirrelli_tex2016_2.pdf

10:30 AM Coffee break

17 Rolls II

Speaker: Prof. Edmund Rolls (University of Oxford)

Paper rolls_paper_tex2016.pdf

Slides rolls_tex2016.txt

12:30 PM Lunch

Contributed talks VI

18 Davis and Adelman I

Speakers: Prof. Colin Davis (University of Bristol), Dr James Adelman (University of Warwick)

Slides davisAdelman_tex2016.pdf

summary Trieste_session_1.docx

5:00 PM Coffee break

19 Davis and Adelman II

Speakers: Prof. Colin J. Davis (University of Bristol), Dr James Adelman (University of Warwick)

summary Trieste_session_2.docx Worksheet_B_SISSA.pdf

Thursday, July 14

20 The blind cognitive scientists and an elephant: Language statistics and perceptual simulation in conceptual processing

Over the last decade the cognitive sciences have strongly advocated an embodied view on cognition. Despite the empirical evidence favoring perceptual simulation in language processing, the conclusions drawn from this evidence often suggest perceptual simulation being the only explanation for conceptual processing. This talk demonstrates that language statistics should not be dismissed in a unified account of cognition. For many findings in the embodied cognition literature attributed to perceptual simulation, language statistics are in fact better predictors, depending on the cognitive task, the nature of the stimulus, individual differences and the time course of processing.

Speaker: Prof. Max Louwerse (Tilburg University)

Paper Louwerse-2011-Topics_in_Cognitive_Science.pdf

10:30 AM Coffee break

21 Davis and Adelman III

Speakers: Prof. Colin Davis (University of Bristol), Dr James Adelman (University of Warwick)

summary Trieste_session_3.docx Worksheet_C_SISSA.pdf

12:30 PM Lunch

22 Special Event 2: Tim Shallice

Speaker: Prof. Tim Shallice (SISSA, Trieste, Italy)

Slides shallice_tex2016.pdf

3:45 PM Coffee break

23 Special Event 3: Chris Mathys

Speaker: Dr Chris Mathys (UCL and University of Zurich)

Social dinner

Friday, July 15

24 Language encodes perceptual information: Evidence from distributional semantics

There is an increasing amount of evidence that language encodes perceptual information. In fact, distributional semantics allow for extracting perceptual relations can be extracted. This talk will give examples of language encoding perceptual information, showing that taking rather simple co-occurrence techniques allow for predicting social networks, valence, iconicity and even longitude and latitude of cities. Moreover, experimental evidence shows that language users rely on these language statistical patterns in their cognitive processes.

Speaker: Prof. Max Louwerse (Tilburg University)

Paper Louwerse-2011-Topics_in_Cognitive_Science.pdf

10:30 AM Coffee break

25 Special Event 4: Yann LeCun

Speaker: Prof. Yann LeCun (New York University and Facebook)

12:30 PM Lunch

Goodbye