close

The Department of Computer Science at the University of Cyprus cordially invites you to the Invited Course Lecture entitled:

How to Learn to Read (If You Are a Machine)

Speaker: Loizos Michael
Affiliation: Open University of Cyprus, Cyprus
Category: Invited Course Lecture
Location: Room 147, Faculty of Pure and Applied Sciences (FST-01), 1 University Avenue, 2109 Nicosia, Cyprus (directions)
Date: Thursday, March 22, 2012
Time: 13.30-14.45 EET
Host: Christos Schizas (schizas AT cs.ucy.ac.cy)
URL: https://www.cs.ucy.ac.cy/colloquium/presentations.php#cs.ucy.pres.2012.michael

Abstract:
If one accepts that the ability to pass the Turing Test is a useful indicator — if not a defining property — of machine intelligence, then one should also accept the importance of endowing machines with the ability to read. But before realizing this goal, one first needs to make more precise what “reading” even means. The task of (Recognizing) Textual Entailment can be seen as capturing a key element of reading: the ability to draw (or at least recognize) inferences from a given piece of text. The original operational definition of the task appeals to human judgment to determine soundness of inference. Such an inherently subjective and anthropocentric definition of the task precludes it, however, from being applicable to machines that acquire the ability to read in an autonomous and highly scalable manner. This talk discusses a definition for textual entailment that retains the formal objective aspects of logical entailment, yet it embraces the statistical nature of the operational definition, and avoids rigidness. The proposed definition views text as a partial depiction of some underlying hidden reality, mapped into a piece of text through the text’s author. Textual entailment is, then, the task of accurately recovering information about this underlying reality. High performance on this task can be provably (under typical assumptions) achieved by training a machine on a corpus of text relevant to the domain of interest, without the need for human supervision to identify the inferences that are expected to be drawn from each piece of text. Experimental results confirm the applicability of the approach in a real-world setting.

Short Bio:
Loizos Michael is a lecturer at Open University of Cyprus (since 2009). He is the founder and director of the Computational Cognition research lab (since 2010), and the academic head of a graduate program of studies in Information Systems (since 2011). Before joining OUC he held a visiting lecturer position at University of Cyprus (2008–2009). He was educated at University of Cyprus, where he received a B.Sc. in Computer Science with a minor degree in Mathematics (2002). He continued his education at Harvard University, where he received an M.Sc. and a Ph.D. in Computer Science (2003 and 2008, respectively). His research focuses on the formal and principled understanding of cognitive processes such as learning and reasoning, and how those are employed by humans and other biological organisms in their everyday lives. Specific areas of interest include: commonsense reasoning, temporal and default reasoning, argumentation, computational learning theory, computational evolution theory, computational story understanding, nature-inspired computation, distributed and multi-agent systems.

  Other Presentations Web: https://www.cs.ucy.ac.cy/colloquium/presentations.php
  Colloquia Web: https://www.cs.ucy.ac.cy/colloquium/
  Calendar: https://www.cs.ucy.ac.cy/colloquium/schedule/cs.ucy.pres.2012.Michael.ics