CUI Graduate Student Seminar

Speaker: Alexis Kauffmann

Date: Thursday, Dec 10, 2009

Time: 12:15pm – 1:00pm

Venue: Room 319, Battelle A

Abstract:

This presentation for graduate students at CUI will be dedicated to machine translation (MT) from French to Japanese and English to Japanese. As Japanese and French have very different syntactic structures, a challenge with MT between those two languages is to avoid generating “unnatural” literal translation, without making any grammar mistake. I will explain the way we try to handle this problem with our grammar-based multilingual translation system “Its-2″, focusing on the example of Japanese “adjectival clause” generation. I will show you some of our results, and compare them with the ones of other MT systems. In the end, I will give a little demonstration of the “Its-2″ MT system.

Postponed: New date to be announced

Speaker: Jana Kludas

Date: Thursday, Nov 26, 2009

Time: 12:15pm – 1:00pm

Venue: Room 319, Battelle A

Abstract:

Benchmarks and evaluation competitions have a long tradition in computer science. For 10 years now, ImageCLEF offers an evaluation platform for image retrieval systems in various domains such as medical retrieval, photo and robotics. The Wikipedia task started in 2008 and attracted a dozen participants each year. Its aim is to investigate ad hoc retrieval approaches in the context of a large and heterogeneous collection of images (similar to those encountered on the Web). Each image is associated with user-generated, unstructured text in English. The task encourages the use of multi modal approaches that investigate the combination of evidence from different modalities. The talk will present the Wikipedia task and collection for 2008/2009 as well as what we can learn from it.

Speaker: Donn Morrison
Date: Thursday, Nov 12, 2009
Time: 12:15pm – 1:00pm
Venue: Room 319, Battelle A

Abstract:
For this graduate student seminar, I will give an overview of CAPTCHA (Completely Automated Public Turing test to tell Computers and Humans Apart) technology and present a system of exploiting CAPTCHAs for annotating images. CAPTCHAs, well known for their utility in the fight against spam, have recently attained recognition as a unique way of channelling the collective, cognitive work of thousands of users onto difficult problems. I propose a novel CAPTCHA system, called TagCaptcha, that exploits the need for human verification on the Web in order to incrementally annotate an image database. I will also present the results from a small usability study and will give a live demonstration of the TagCaptcha system.
Demo: http://dolphin.unige.ch/tagcaptcha/
Presentation slides