The '02/'03 parlepraatjes

In mijn afstudeeronderzoek wil ik het hierboven beschreven model realiseren in een (waarschijnlijk object-georienteerde) architectuur, waarbij (in ieder geval) de presentatielaag gegenereerd wordt uit een declaratieve beschrijving in een XML-variant (misschien UIML). Ook ben ik geinteresseerd in het automatisch enforcen van complexe constraints binnen de conceptuele laag (bijvoorbeeld, bij het beschrijven van een datum: ALS maand = april DAN dag <= 30).

Samengevat: user interface programmeren/modelleren op een wat hoger niveau van abstractie, en zo min mogelijk imperatief.

Virtuele Persoonlijkheden: we kunnen deze maken, maar hebben ze een toekomst?

Al-Noor Ladhani

Presenting a Monument in Restoration: The Saint Laurentius Church in Ename and Its Role in the Francia Media Heritage Initiative

Kris Forster

Virtual Reality and multimedia technologies are central components of the heritage presentation programme at Ename, Belgium. These techniques are designed to help the visitor understand and experience the past as it has been revealed through archaeological and historical research. This paper briefly traces the development of multimedia and Virtual Reality technologies in presenting the archaeo-logical site of the early medieval fortress and St. Salvator Abbey at Ename and in the archaeological exhibits in the Ename Provincial Museum. It then highlights the approach used in Ename to present the 10th-century Saint-Laurentius Church and its restoration to the public, describing the methodology of creating scientifically verifiable Virtual Reality reconstructions. In conclusion it will place this project in the context of an international heritage initiative, the Francia Media Project, in which new interpretive tech-nologies will be developed by members of a scientific con-sortium to offer new media of public heritage presentation in a European context.

TALUS

Marc Blasband

In 2002 we returned to the TALUS platform to integrate new ideas with established concepts for language recognition and dialog management:

• A structure with topics reduces the complexity and the quantity of homonyms (when the same word has two meanings, for example bank or plain).
• The users' world-view can be captured through language.
• Language Processing is most meaningful when complex applications are being developed.
• A dialogue with only two levels is close to the way of thinking of most people.
• Learning and teaching paradigms allow dynamic creation of dictionaries and grammars.
• An off-the-plank dictation system used as front end handles one part of the problem.
• A complete application is more convincing than parts of it.

Syntactical analysis of natural languages using Dependency Grammar

Vidas Daudaravicius

Context-Free grammars are used with reasonable success for parsing many languages (e.g. English, French, Spanish, German, Dutch, Chinese ...). In all these languages, word order is important for the meaning of a sentence. For example, the meaning of "the dog bites the man" is different from the meaning of "the man bites the dog". This word order is used in an essential way in context-free grammars.

On the other hand, context-free grammars can not be used for parsing inflectional languages (e.g. all slavic, baltic languages). One distinguishing feature of such languages is that they have so-called "free word order", i.e. the order of the words within a sentence can often be changed freely without changing the meaning of the sentence. These languages use inflections of words to determine the meaning of a sentence.

An acceptible approach to parsing sentences in inflectional languages is by means of Dependency Grammars. In these grammars words "depend on" other words in the same sentence, thus giving rise to different grammatical structures than context-free grammars. Dependency grammars are in a way more semantical in nature. They can be succesfully used for functional languages too. I will discuss Dependency Grammars, and apply them to Lithuanian, a language which has free word order.

MUMIS: annoteren (tool) en grafisch representeren

Galina Slavova en Niek Schmoller

In verband met het Europese MUMIS project ontstond de behoefte voetbalverslagen handmatig te "annoteren". Met het annoteren worden gebeurtenissen op een bepaald tijdstip van een wedstrijd opgeslagen in . Deze annotatie kan later door de computer verder verwerkt worden, bijv. om te zoeken in archieven.

Om het annoteren makkelijker te maken hebben wij een "Annotation Tool" gemaakt.

Daarnaast hebben we nog een tweede opdracht uitgevoerd, namelijk het grafisch representeren van een boomstructuur van een grammaticaal ontleedde zin.

Het VMC in Java3d, ingeladen vanuit X3D bestanden

Ruud Greven

Het virtuele muziekcentrum (VMC) is herbouwd, zodanig dat het makkelijk uit te breiden is, en gebruikt kan worden in combinatie met Java. Er is gekozen om het eerst om te zetten naar X3D en dat vervolgens in te laden met een loader. In deze presentatie zal ik gaan vertellen over deze X3D loader.

Ik zal eerst vertellen welke werkzaamheden ik heb verricht tijdens mijn stageperiode hier, en welke stappen ik heb genomen om tot een goed eindresultaat te komen. Hierbij zal ik ook ingaan op problemen die ik tegen ben gekomen, en hoe ik die opgelost heb. Tevens zal ik kort vertellen hoe de X3DLoader intern in elkaar zit en hoe hij is uit te breiden met nieuwe functionaliteit. Aan het einde zal nog een korte demonstratie komen van de mogelijkheden op dit moment.

Animating Embodied Agents

Niels Duijn en Marnix van Ommen

Gedurende onze afstudeerstage hebben wij verschillende manieren ontwikkeld om een avatar te kunnen animeren. Hierbij is de nadruk gelegd op het maken en aansturen van bewegingen, waaronder vooral een aanwijsbeweging. Het doel is om deze bewegingen in het virtuele muziekcentrum (VMC) toe te passen.

Technieken die we gebruikt en / of onderzocht hebben zijn: forward en inverse kinematics, X3D, Java3D en animatiescripttalen.

Ontwikkeling van een Spaans electronisch woordenboek voor Cubaanse scholieren.

Koen Meijs

Gedurende mijn stage in Santiago de Cuba heb ik meegewerkt aan de ontwikkeling van een Spaans electronisch woordenboek bedoeld voor Cubaanse scholieren. In de presentatie zal ik een overzicht geven van het complete ontwerpproces van het woordenboek en technieken die daarin toegepast zijn. Daarnaast zal ik de methode van spelfoutdetectie die we hebben toegepast nader toelichten.

Affective experiences: designing and evaluating emotions

Claire Dormann

Much of the human-computer interaction (HCI) community has been limited for years to measuring and responding to aspects of human behaviour such as efficiency or capacity, in this reflecting HCI major concern with productivity. However, the use of computer technology has reached far beyond the office toward the home and community. Emerging computer applications elicit a variety of emotions ranging from joy or fear, to trustworthiness or sophistication. ), the emotion of trustworthiness has been emphasised, as one of the most important factors in electronic commerce. These applications can generate social emotions (e.g. familyware) or emotional reactions (e.g. interactive games). They can also be an expression of emotional self-awareness (e.g. emotion container), or an object of emotions (e.g. interactive toys). To create affective experiences and measure users' reactions, require further development of methodologies, the operationalisation of several new variables and an extended conceptual framework.

In this talk, I will focus on affective experiences in relation to agent and electronic commerce. A small case study in the role visual humour will be briefly discussed. Then I will move the evaluation of emotions, presenting methods for assessing emotions. Emotional evaluation will be related to on-going work in agent evaluation. The purpose of this talk is to present ideas for the development of an affective framework and outlined issues for future research.

Knowledge Mapping

Willem-Olaf Huijsen

Knowledge mapping is the field within knowledge management that aims to optimise the effective and efficient use of the corporate knowledge base by addressing the following issue of how can one best support finding the knowledge that is available within an organisation, and building insights into the qualities of this knowledge.

Knowledge-mapping tools provide computer support for addressing these issues. These tools may interface with any number of information systems to collect data on the knowledge that is available within the organisation. A knowledge-mapping tool typically contains a map of the expertise of the organisation in terms of a set of concepts. This may be a simple list of keywords, a more complex thesaurus with synonymy and other basic relations between concepts, or even sophisticated ontologies with lots of domain knowledge encoded into them. The tools may offer various views on the collected data, with customisation to different organisational roles and personal preferences.

As part of the presentation, a prototype of a knowledge-mapping tool will be demonstrated.

Expressing Emotions: visual capabilities of two 3D talking heads

Niels Besseler

During my internship at ITC-irst in Povo (Italy) I have been evaluating two talking heads' abilities to express emotions with their faces. The results of this evaluation will be used in ITC-irst's research on emotional talking heads and will possibly be connected with computational humour experiments.

In this talk I will present a short overview on the work I have done and the technologies involved. The choices I made and results I found will be discussed briefly, and to conclude the talk, some example experiments with the talking heads will be shown.

GRETA. Facial animation and embodied conversational agents

Catherine Pelachaud, Paris 8

Our goal is to develop a Believable Embodied Agent able to dialog with a user, but we aim at making an Agent that can also combine facial expressions in a complex and subtle way, just like a Human Agent does. We are applying the metaphor of face-to-face communication to human- computer interaction.
Face-to-face conversation is very complex phenomenon as it involved a huge number of factors: we speak with our voice, but also with our hand, eye, face and body. Our gesture modifies, emphasizes, contradicts what we say by words. It is therefore important to consider both verbal and nonverbal behaviors while building an embodied agent. The agent must have the capacity to decide which facial expressions to show, which words to say with which intonation.
That is, the agent should be able to plan not only what to communicate, but also by what (verbal or nonverbal) signals, in what combination and how synchronized.

In this talk, we will present our work on the creation of a multimodal believable agent able to dialog with a user and whose nonverbal behaviors reflect its affective state.

CV
Catherine Pelachaud received her PhD in Computer Graphics at the University of Pennsylvania, Philadelphia, USA 1991, under the supervision of Norman Badler, and Mark Steedman.

She has contributed to the MPEG-4 standard for facial animation.

She has held positions at the University of Rome at the computer science department. She recently moved to Paris 8.

Tailoring from a User Perspective

Margit Biemans (Telematica Instituut)

At this moment, more and more systems are designed tailorable to meet dynamic and diversified requirements. Tailoring means that users can adjust the system after implementation, and that such adjustments are persistent across sessions. Examples of tailoring are changing the ring tone of your mobile phone, adding a filter to your e-mail and creating a Macro in Word. Evaluation studies show that tailoring is not used to its full potential. Therefore, the research question of my PhD research is: How to support users in beneficial tailoring?

In order to answer this question, we took the following approach. First, we conducted some explorative evaluation studies to identify tailoring habits and tailoring problems. Next, we designed a theoretical framework that describes and prescribes normal tailoring behaviour, and guides the interaction design of tailorable applications. Finally, this theoretical framework will be validated in user experiments.

In this presentation, the focus is on the theoretical framework of tailoring. As no single tailoring theory is available yet, we selected and fine-tuned some existing (behavioural) theories, i.e., Information Foraging Theory, Bounded Rationality, Situation Theory, Channel Theory and a Social Theory on Accretions, Tuning and Restructurings. The use and the relation between these theories are illustrated with some examples from a medical tele-consultation setting.

The eventual results of this PhD research are an improved understanding of tailoring behaviour, and central guidelines (or tips) to improve the usability of tailorable systems. These latter ones are to transferable to future applications like location-based, context-aware and ubiquitous services.

Learning of Prediction Strategies

Mark van Setten (Telematica Instituut)

Na een korte inleiding over zijn onderzoek naar 'Prediction strategies for recommender systems' wil Mark een aantal inhoudelijke problemen bespreken in de hoop zinvolle/nuttige suggesties te krijgen. Dit pdf-bestand vat het onderzoek en de problemen kort samen.

Improvements on a simple muscle-based 3D face for realistic facial expressions

The Duy Bui

Facial expressions play an important role in face-to-face communication. With the development of personal computers capable of rendering high quality graphics, computer facial animation has produced more and more realistic facial expressions to enrich human-computer communication. In this talk, we present a simple muscle-based 3D face model that can produce realistic facial expressions in realtime. We extend Waters' muscle model to generate bulges and wrinkles and to improve the combination of multiple muscle actions. In addition, we present techniques to reduce the computation burden on the muscle model.

YPS - The Ynperfect Pun Selector for Computational Humor

Christian F. Hempelmann
Purdue University, West Lafayette

One of the major concerns about humor in intelligent agents is their ability to go beyond recalling and citing databased material, such as offering to tell a joke apropos a text being entered on the computer. This involves various forms of generating new material.

My current research describes a method to evaluate and select phonologically possible and better imperfect puns for use in computational on-the-fly pun generation in human-computer interfaces (HCI). Its result is aimed specifically at a module to generate such imperfect puns, in short and for the sake of a memorable acronym: an YPS (ynperfect pun selector).

YPS is intended as a complement to a general pun generator that forms part of humor component, based on the General Theory of Verbal Humor (GTVH, Attardo and Raskin 1991), itself to be integrated into a full-fledged natural language processing (NLP) system based on ontological semantics (Nirenburg and Raskin 2003).

While the focus is on the imperfect pun selector, the humor module and necessary adaptations to such an NLP system are outlined in an effort to conceptually integrate the YPS module into future efforts at computational humor, i.e., its reusability.

Alternating-time Temporal Epistemic Logic

Wojtek Jamroga

Several formalisms based on temporal logic have been proposed recently to tackle verification of multi-agent systems properties. I would like to address two of them: Alternating-time Temporal Logic (ATL) and Alternating-time Temporal Epistemic Logic (ATEL). ATL was introduced by Alur, Henzinger and Kupferman in the late 90s; it's based on the standard branching-time temporal logic CTL, but with a game theory twist (notions of players, strategies, game structure etc.). In consequence, it offers some nice ways of expressing capabilities and temporal properties of autonomous agents and their coallitions. ATEL, proposed by van der Hoek and Wooldridge in 2002, enriches the picture with an epistemic component to reason about agents' beliefs.

There are still problems with ATEL semantics, though. The semantics proposed in the original paper by Mike Wooldridge and Wiebe van der Hoek is well founded in the mathematical sense, of course, but it implies properties that are counter-intuitive. I thought I proposed a good solution to this proplem in my FAMAS paper - but Wiebe van der Hoek has already pointed out that my proposal is not completely correct either. It seems that the problem is more difficult than everybody had expected at the beginning.

The semantics of ATEL is still a subject of ongoing research. We have managed to clarify a few issues, although the number of questions grows rather than decreases. In my talk, I would like to present the idea of both ATL and ATEL, our attempts at ATEL semantics (and why they failed), the research questions we are considering now, and the solutions we seem to be approaching.

Web crawling and data mining for layout aspects of web pages

Stefan de Bruijn

In this Parlepraatje, Stefan de Bruijn will present the results of two 'vrije projecten' he has carried uit under supervision of Eelco Herder and Betsy van Dijk.

With the growing importance of the Internet, the need for designing a quality website becomes more clear. Besides the site content and site structure, layout is an important aspect as well. These observations have led to the starting point of these projects: will it be possbile to extract from existing web sites a 'most appreciated' layout, or at least one that draws attentions.

There is no straightforward answer to this question. To start with, how important are the web site's contents? And how do we rank a web site? Does layout depend particularly on the category of topics that dealt with? In the first part of this talk, Stephan will present a web crawler that has been specifically designed for this purpose. Data analysis will be dealt with in the second part, followed by the conclusions drawn.

Grouplens Research Project

Sean M. McNee

Over the past decade, recommender systems employing collaborative filtering technology have evolved from research proofs-of-concept to commonplace components of e-commerce web sites and direct marketing. At the same time, research has moved forward to embrace more sophisticated algorithms and more detailed exploration of the user experience. This talk reviews recent highlights from the GroupLens Research Project, including an exploration of techniques for getting new users started in recommender systems, building confidence into a recommender system, user interface design issues, using citation webs to recommend research papers, and recent advances in algorithm design.

Sean M. McNee is a doctoral candidate from the Department of Computer Science and Engineering at the University of Minnesota. As a member of the GroupLens Research Project, he performs research in the areas of recommender systems, user modeling, and human-computer interaction. He received his M.S. in Computer Science from the University of Minnesota in 2003, and his B.A. from Carleton College in Northfield, Minnesota in 1998.

• Parlepraatjes
• Previous praatjes
• 08/'02 - 06/'03 praatjes (this page)
• 08/'01 - 06/'02 praatjes
• 08/'00 - 06/'01 praatjes
• 10/'99 - 06/'00 praatjes
• 10/'98 - 06/'99 praatjes

Last modified at $Date: 2003/09/18 14:05:48 $ by Hendri Hondorp