Maria-Hendrike Peetz: "Cognitive Temporal Document Priors" and "Online Reputation Management"

Ponente: Maria-Hendrike Peetz, (Information and Language Processing Systems, UvA)

Fecha: martes 19 de marzo de 2013

Hora: 12h00

Lugar de celebración: Sala 1.03, ETSI Informática, UNED

Abstract

Cognitive Temporal Document Priors

Temporal information retrieval exploits temporal features of document collections and queries. Temporal document priors are often used to adjust the score of a document based on its publication time. We consider a class of temporal document priors that is inspired by retention functions considered in cognitive psychology; such functions are used to model the decay of memory. We find that many retention functions used as a temporal document prior have a positive effect on overall retrieval performance. We examine the stability of this effect across news and microblog collections and discover interesting differences between retention functions. We also study the problem of optimizing parameters of the retention functions as temporal document priors and again find interesting differences where some retention functions display consistent good performance across large regions of the parameter space. Overall, a retention function based on a Weibull distribution is the preferred choice for a temporal document prior.

Online Reputation Management

My current research interest lies in Online Reputation Management, in particular the estimation of reputation polarity. I will explain my current approaches and plans.

Bio

Hendrike is a PhD student at the University of Amsterdam, at the Information and Language Processing Systems (ILPS) group at University of Amsterdam. She works on temporal information retrieval for social media. She's currently spending some weeks at NLP&IR-UNED Group as a visiting researcher.

Lugar de celebración

Sala 1.03
ETSI Informática, UNED
c/ Juan del Rosal, 16
Ciudad Universitaria
28040 Madrid

 
RocketTheme Joomla Templates