Invited Talks

Speaker & Bio


Gina R Kuperberg

Gina R Kuperberg, MD PhD


Monday morning, June 10th

Predicting Meaning: What the Brain tells us about the Architecture of Language Comprehension

It is well established that we draw upon our real-world knowledge to predict upcoming events and even individual words. I will discuss evidence that the neurocognitive mechanisms that we engage in retrieving conceptual information associated with incoming words are quite distinct from those engaged when these predictions are disconfirmed by the input. Drawing broad links with computational models conceptualizing language comprehension as an incremental process of belief updating, I will suggest that the engagement of these distinct neurocognitive systems allows for comprehension that is both highly efficient and highly flexible (1).

I will first discuss studies using event-related potentials (ERPs) to examine online brain activity during sentence and discourse comprehension. I will then draw some (still tentative) links between this ERP literature and some relevant fMRI and MEG studies. Finally, I will discuss the advantages of a predictive comprehension system. Predicting correctly clearly offers advantages in terms of computational efficiency. Here I will argue that the costs incurred when we predict incorrectly are also crucial for successful and flexible comprehension. Neurocognitive responses triggered by prediction errors may rescue us from interpretation errors in noisy environments, may allow us learn novel events, and may enable us to flexibly adjust our comprehension strategies in response to everchanging task and environmental demands.

Kathleen McKeown

Kathleen McKeown, PhD


Wednesday morning, June 12th

Natural Language Applications from Fact to Fiction

Much research in the natural language field has been carried out on news, given the large amount of annotated data now available for this genre.

Certainly, the ability to analyze the facts of real world events, enabling systems to answer questions and summarize the events of the day is an important application. As research has moved to analysis of new genres, whether fact, opinion or fiction, new approaches to existing applications have arisen and opportunities for new applications have emerged. In this talk, I will present research in my group at Columbia as it has moved from news to scientific articles to online discussion forums to novels. I will touch on summarization, open-ended question answering, social influence and social networks.