BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Towards More Robust and Interpretable Models for Structured Predic
 tion and Language Generation - Adhiguna Kuncoro\, DeepMind
DTSTART:20171020T110000Z
DTEND:20171020T120000Z
UID:TALK89091@talks.cam.ac.uk
CONTACT:Amandla Mabona
DESCRIPTION:Many important tasks in natural language processing involve pr
 oblems with structured output spaces\, such as machine translation and syn
 tactic parsing. In line with recent advances in representation learning\, 
 I will describe two ways to improve state of the art neural methods in lan
 guage modeling and syntactic parsing. Despite the expressive power of recu
 rrent neural networks\, I explore the hypothesis that such models can stil
 l benefit from linguistically-motivated inductive biases\, thereby facilit
 ating better generalization given a limited amount of labeled data. I will
  also argue for the importance of cost functions in learning\, where a ref
 ined notion of costs\, based on the idea of ensemble distillation\, improv
 es the performance of a strong neural dependency parser. \n\nIn the opposi
 te direction\, neural representation learners can be thought of as “mini
 -linguists”\, in the sense that they often need to come up with generali
 zations and theories about certain latent aspects of language. The learner
 s’ findings can then be used as empirical means to confirm or refute cer
 tain linguistic hypotheses. To this end\, I demonstrate that the findings 
 of Recurrent Neural Network Grammars (RNNG)\, a state of the art model for
  parsing and language modeling\, mostly align with certain linguistic theo
 ries of syntax\, while also discovering some syntactic phenomena that are 
 different from our intuition\, but are interesting nevertheless.\n\nBIOGRA
 PHY:\nAdhiguna Kuncoro is a DPhil student in computer science at the Unive
 rsity of Oxford and research scientist at DeepMind. His primary research i
 nterest lies at the intersection of natural language processing and machin
 e learning\, particularly on designing statistical models of natural langu
 age that are: i.) robust\, ii.) interpretable\, and iii.) able to learn mo
 re with less amount of annotated data. His co-authored work received an Ou
 tstanding Long Paper award at EACL 2017. He holds a Master’s degree in l
 anguage technologies from CMU LTI\, where he worked on low-resource langua
 ge processing under the DARPA LORELEI project\, and another Master’s deg
 ree in computer science from the University of Oxford. His DPhil study at 
 Oxford is jointly funded by an EPSRC studentship under the flexibility to 
 support the very best students scheme\, the Oxford Department of Computer 
 Science scholarship\, and a Balliol Mark Sadler scholarship award. He comp
 leted his undergraduate degree in Informatics Engineering from Institut Te
 knologi Bandung\, Indonesia.
LOCATION:FW26\, Computer Laboratory
END:VEVENT
END:VCALENDAR