Vés enrere 10 i 15/11/17: Conferències de recerca a càrrec de Marco Baroni i Jason Rothman

10 i 15/11/17: Conferències de recerca a càrrec de Marco Baroni i Jason Rothman



En el marc del procés de valoració de candidats per ser avalats per a la convocatòria ICREA, els dies 10 i 15 de novembre hi haurà les presentacions de recerca de Marco Baroni i Jason Rothman, respectivament. La informació sobre aquestes presentacions és la següent:
Títol: Spectacular successes and failures of recurrent neural networks applied to language
A càrrec de: Marco Baroni
Data: divendres 10 de novembre del 2017
Hora: 12.00 - 13.00 h
Lloc: sala de reunions 52.701
Recurrent neural networks (RNNs) are attractive computational models to test hypotheses about language acquisition and use, due to their generality and ability to induce linguistic knowledge from raw data. In this talk, I will first report the results of an experiment suggesting that RNNs are extracting surprisingly abstract grammatical generalizations from corpora (they correctly predict that "the colorless green ideas you slept yesterday" should continue with a verb in the plural form). I will then report a second experiment suggesting that RNNs are not "systematic" in Fodor's sense: even when trained to process the commands "run", "run twice" and "jump", at test time they fail to correctly predict the meaning of "jump twice". Finally, I will outline my research agenda for the next few years, aiming at equipping RNNs with *compositional* skills, in order to endow them with more human-like capabilities to acquire and use natural language.
Títol: Different? Probably. Incomplete?  Absolutely not: Tracing sources of differences in heritage bilingual outcomes 
A càrrec de: Jason Rothman
Data: dimecres 15 de novembre del 2017
Hora: 16.30 - 17.30 h
Lloc: sala de reunions 52.701

In this talk, I will first introduce the audience to and problematize both the concept of what a heritage language bilingual is and the literature that has studied their competence outcomes in adulthood over the past two decades.  Heritage speakers are native–often child L1 or 2L1– speakers of a minority “home” language who (usually) become dominant speakers starting at school-age in the societal majority language of the (national) community in which they grow up and are educated.  Typically, heritage speakers show interesting differences in their knowledge and performance in the heritage language as compared to age-matched monolinguals.  Often, such differences have been labelled as instances of incomplete acquisition (e.g. Montrul 2008, 2016) or attrition (Polinsky 2011, 2015).  Under both accounts, although for different reasons, heritage language bilingual differences are viewed as some type of deficiency.  I will propose that many differences, alternatively, could have only developed the way we see them in heritage grammars for reasons related to qualitative differences in the input heritage speakers receive (e.g. Rothman 2007; Pires and Rothman 2009;  Pascual y Cabo and Rothman 2012; Kupisch and Rothman, 2016).  In doing so, I will link a process of cross-generational attrition and access to literacy in the heritage language to (some) outcomes in heritage language development. I conclude by suggesting that many aspects argued to be incompletely acquired in heritage language grammars are in fact complete, but unavoidably different.


Animem tots els membres de l'UCA a assistir a aquestes presentacions. A més, tothom que vulgui pot reunir-se amb els professors Marco Baroni i Jason Rothman per conèixer millor la seva recerca i valorar el seu possible encaix a l'UCA: només cal enviar un missatge a [email protected] indicant amb quin dels dos professors es voldria reunir (o amb tots dos, si és el cas) i programaré les reunions que es demanin. 



ODS - Objectius de desenvolupament sostenible:

Els ODS a la UPF