The 29th UW/MS Symposium in Computational Linguistics

Опубликовано 27 июля 2016, 0:44

Title: Learning with Weak Supervision in Grounded Language Acquisition Speaker: Hannaneh Hajishirzi A central problem in grounded language acquisition is to learn the correspondences between a rich set of events and complex sentences that describe those events. In this talk, I will introduce a novel approach to learn these correspondences under weak supervision that comes from loose temporal alignments between events and sentences. The core idea is to exploit the underlying structure between correct, but latent, correspondences using a discriminative notion of similarity coupled with a ranking function.This algorithm reasons in terms of pairwise discriminative similarities and utilizes popularity metrics to learn the alignments between events and sentences and even discover group of events, called macro-events, that best describe a sentence. I will demonstrate extensive evaluations on our new dataset of professional soccer commentaries. Furthermore, I will describe how this model can be applied under the general framework of Multiple Instance Learning. Title: Mining Online Social Behavior for Enhanced Behavioral Health Speaker: Munmun De Choudhury People are increasingly taking on to social media platforms, as Twitter and Facebook, to share their thoughts and opinions with their audiences. Often these updates are made in a naturalistic setting over the course of daily activities and happenings. Beyond elucidating core aspects of how we act, interact or emote, these platforms thus provide a promising mechanism to capture behavioral attributes relating to an individual�s social and psychological environment, some of which may signal concerns about their mental health. In this talk, we will examine the harnessing of social media as a tool in behavioral health. Today affective disorders constitute a serious challenge in public health: More than 9% of US population is known to suffer from depression. I will present two problems where social media behavior can help us reveal latent affective concerns. First, I will discuss the use of social media, particularly online activity, emotion and linguistic expression, in making inferences about behavioral changes in new mothers following childbirth. Second, I will present predictive models that leverage social media behavioral cues, to detect, ahead of onset, the likelihood of major depression in individuals. Broadly, such predictive forecasting can help develop unobtrusive diagnostic measures of behavioral disorders, and can hopefully enable behavioral health tracking and surveillance in large populations in a fine-grained manner. I will conclude with the potential of this line of research in informing the design of next generation low-cost, privacy-sensitive early-warning systems and interventions, that can bring people timely information and assistance.