2007 Conference Program


< Back


Sunday, Jan 28 Monday, Jan 29 Tuesday, Jan 30 Wednesday, Jan 31
Tutorials and Workshops 8:45 - 9:00
Welcome

9:00 - 10:30
Papers: User Modeling
9:00 - 10:30
Invited Speaker: Thomas Strothotte
9:00 - 10:30
Invited Speaker: Susan Dumais
10:30 - 11:00
Break
10:30 - 11:00
Break
10:30 - 11:00
Break
11:00 - 12:30
Papers: Recommender Systems
11:00 - 12:30
Papers: Information Retrieval
11:00 - 12:30
Papers: Natural Language Interfaces
12:30 - 2:30
Lunch
12:30 - 2:30
Lunch
12:30 - 2:30
Lunch
2:30 - 4:00
Papers: Social Software
2:30 - 4:00
Papers: Personal Assistants
2:30 - 4:00
Papers: Multi-Modal Interfaces
4:00
Luau
4:00 - 4:30
Break
4:00 - 4:30
Break
4:30 - 5:30
Papers: Demonstration Based Interfaces
4:30 - 6:00
Papers: Gesture-and-Sketch-Based Interfaces
Opening Reception 5:30 - 6:00
Poster Preview Session
7:00 - 9:00
Poster Session
Farewell


Tutorials and Workshops
Sunday, January 28th

 


T1: Building Ubiquitous and Robust Speech and Natural Language Interfaces (ACM Digital Library Link)

Gary Geunbae Lee
Professor
Isoft Lab, POSTECH
Pohang, Korea

Shimei Pan
Research Scientist
IBM T. J Watson Research Center
Hawthorne, NY USA

Benefits
Attendees will come away with an in-depth understanding of principles and technologies that can be used in building robust and ubiquitous speech and natural language interfaces including spoken dialog systems, speech-to-speech translation systems and intelligent chatting systems

Origins
New for IUI 2007

Features
The full day tutorial consists of the following two parts.

1st half day (morning): Technologies for ubiquitous speech and natural language user interfaces

  1. Introduction to ubiquitous spoken dialog systems
  2. Automatic speech recognition and post-processing
  3. Spoken language understanding and dialog management
  4. Clarification dialog and dialog development studio
  5. Intelligent chatting and speech-to-speech translation
  6. Summary of ubiquitous spoken dialog systems

2nd half day (afternoon): Building robust speech and natural language user interfaces

  1. Analyses of problems in speech and natural language user interfaces
  2. Typical error detection strategies
  3. Typical error recovery strategies
  4. Multimodal error correction
  5. Adaptive speech and natural language interfaces
  6. Summary of design principles

Audience
Researchers, students, designers and application developers who wish to incorporate natural language interactions and dialog techniques in their systems.

Presentation
Lecture and demonstration

Instructors
Gary Geunbae Lee has been a professor at CSE department, POSTECH in Korea since 1991. He is a director of Intelligent Software (ISoft) Laboratory which focuses on human language technology researches including natural language processing, speech recognition/synthesis, speech translation, question answering and web/text mining. Professor Lee authored more than 100 papers in international journals and conferences, and has served as a technical committee member and reviewer for several international conferences such as ACL, COLING, IJCAI, ACM SIGIR, AIRS, ACM IUI, Interspeech-ICSLP/EUROSPEECH, EMNLP and IJCNLP. He is currently leading several national and industry projects for robust spoken dialog systems, spoken dialog translation and expressive TTS. Professor Lee holds a Ph.D. in computer science from UCLA, and BS/MS in computer engineering from Seoul National University.

Shimei Pan has been a research scientist at the IBM T.J. Watson Research Center in New York since 2000. She is a member of the Intelligent Multimedia Interaction Department, which focuses on developing natural language and graphics interaction technologies to facilitate robust and effective human-computer interactions. Currently, she is working on developing new methods that can improve the robustness and usability of practical conversation systems. Previously, she has been working on case-based natural language generation, intelligent navigation for conversation management, prosody modeling for spoken utterance generation, user preference modeling, and multimedia coordination and synchronization for multimedia presentation generation. She holds a PhD in computer science from Columbia University.

[top]


T2: Advanced Topics in Recommendation (ACM Digital Library Link)

John Riedl
Professor
Department of Computer Science
CS Building
200 Union Street SE
Minneapolis, MN 55455
U.S.A.
http://www-users.cs.umn.edu/~riedl/

Anthony Jameson
Principal Researcher
DFKI, German Research Center for Artificial Intelligence
Stuhlsatzenhausweg 3
66123 Saarbrcken
Germany
http://dfki.de/~jameson/

Benefits
Attendees will come away with an up-to-date, active understanding of a representative set of current developments in recommender systems that will help them to conduct cutting-edge research and/or to work more effectively with the currently widespread recommendation technology.

Origins
The instructors have previously (co-)presented tutorials on recommender systems in the IUI, CHI, IJCAI, AAAI, EC, CSCW, and UM conference series. But the material in this tutorial is largely new, because of the focus on advanced topics from recent and current research.

Features
Recommender systems have been one of the most successful forms of intelligent user interface. Despite the widespread commercial deployment of now-familiar, relatively lightweight forms of recommendation, researchers have been exploring new paradigms and issues that should allow recommendation to be applied in new application settings and with increasing user acceptance.

We will begin with a brief recapitulation of well-known concepts and techniques, so that even IUI 2007 attendees who have little prior familiarity with recommender systems will be able to understand the rest of the tutorial. About 4/5 of the tutorial will be devoted to a selection of advanced topics that seem likely to take on increasing importance in research and practice in the near future, such as the following topics:

  1. Recommendation for groups of users: When a recommendation is generated not for an individual but for a group of persons that will act on the recommendation together, the recommendation problem takes on new dimensions.

  2. Recommenders and the social web: The social web is changing our thinking about the Internet: no longer just a place to read and write, it is becoming a place to meet, chat, argue, discuss, confess, and befriend. We will examine the ways recommenders are helping to create and nurture the social web.

  3. Recommendations of coherent sets of items: Often a user will be most interested in recommendations that span a space, rather than a narrow set of recommendations all of the same type. For instance, recommending a set of technical papers all by the same author may be less interesting than recommending a range of interesting papers. Recommender algorithms must be extended to incorporate a notion of the separation among items in a set.

  4. Recommendations concerning sequences of actions: In some settings it is natural to recommend actions that together will form a sequence (e.g., visits to stores in a shopping center), where the choice of the later actions will depend on the results of the actions earlier in the sequence (e.g., whether a desired product was found in the first store visited). This paradigm requires different recommendation techniques and raises new interface design issues.

  5. Question-asking in recommenders: How can recommenders discover a sequence of questions to help a user select an item that fits his or her needs? In many applications the user will not be aware of the sets of available options or of the mapping between the options and user needs. The recommender can help the user explore the item space with appropriate questions based on the answers to previous questions.

  6. Exploitation of the users episodic memory: The increasing technological possibilities for automatically capturing a users actions and experiences (e.g., prior encounters with relevant products in various settings) increases the range of data that can be used for recommendation, as well as the variety of ways in which recommendations can be presented (e.g., in conjunction with reminders of past experiences).

  7. Security and recommenders: How do recommenders relate to security and privacy? What can be done to protect recommenders from malicious or accidental attack?

Audience
Researchers, students, designers, and application developers who are working to improve recommendation technology and/or the associated user interfaces. In particular, even participants who do not conduct advanced recommender research themselves should be able to benefit from a better understanding of the alternatives to current technologies and application settings.

Presentation
Each topic will be introduced and motivated with examples from recent research and/or practice. The new recommendation methods will be presented and discussed critically, and open issues will be identified.

Because this is a full-day tutorial, enough time will be allocated to each topic to allow interactive discussion with the participants as well as the inclusion of some brief exercises designed to ensure active understanding.

Instructors
John Riedl has been a member of the faculty of the Computer Science department of the University of Minnesota since March 1990. In 1992 he co-founded the GroupLens Research project on collaborative information filtering, and has been co-directing it since. In 1996 he co-founded Net Perceptions to commercialize GroupLens. Net Perceptions was the leading recommender systems company during the Internet boom. In 1999, John and other Net Perceptions co-founders shared the MIT Sloan Schools award for E-Commerce Technology. Recently, GroupLens Research has been exploring the ways in which recommender systems can be used to study and enhance social structures, such as conversation systems. Can the recommender help members understand their connections to the group in deeper ways? How about each members unique contributions to the group? Or, the value of a members contribution to the success of the group as a whole?

Anthony Jameson is a principal researcher at DFKI (the German Research Center for Artificial Intelligence) and adjunct professor of computer science at the International University in Germany. Since the 1980s, he has been among the first researchers to work on various paradigms in recommendation, including conversational recommendation, recommendation based on multiattribute utility theory, probabilistic estimation of users evaluation criteria, recommendation of sequences of actions, and the integration of recommending and reminding. He has also published widely on other topics in intelligent user interfaces, including systems that adapt to their users and natural language dialog systems.

[top]


T3: Designing Intelligent User Interface for Ubiquitous Computing Environments (ACM Digital Library Link)

Dr. Anxo Cereijo Roibas
Senior Lecturer
University of Brighton, UK
A.C.Roibas@bton.ac.uk

Antonio Krger
Associate professor for Geoinformatics and Computer Science
University of Mnster Germany
kruegera@uni-muenster.de

Abstract
This full-day tutorial will provide an overview on Intelligent User Interfaces in the context of the technological development associated with the Ubiquitous Computing paradigm. Based on a technological assessment of the main technological areas involved (miniaturization, communication and network structures, advanced display and speech technologies, as well as sensor integration and deployment) this tutorial will discuss the main approaches in context and user modeling, as well as present new interaction paradigms for ubiquitous computing environments. Examples of past and current research projects of the field will be used to explain and discuss the intersection of Ubiquitous Computing and Intelligent User Interfaces. Having discussed the technical foundations, the tutorial aims to explore the design and development of new pervasive applications for mobile devices and other distributed interfaces. It will explore how handhelds can interact with other surrounding devices, intelligent environments and with the social context. The tutorial will also look at how innovative ethno-methodologies such as living labs, on-the-field data collection, enactments, 'Cultural Probes', Participatory Design approaches and advanced in-situ evaluation techniques can lead to the creation and representation of feasible and relevant future communications scenarios.

Target Audience
The tutorial aims at graduate students, teachers and professionals that would like to obtain an overview on the major technological developments in the area of Ubiquitous Computing and the major challenges for the design and the development of Intelligent User Interfaces

Structure
This will be a full-day tutorial (2 x 3 hrs) structured into six sessions of one hour each. Session one will discuss the technological trends and developments in the area of Ubiquitous Computing, session two will introduce the major approaches to model contexts and users for adaptive systems in instrumented environments, session three will provide an overview on past and current research projects in the area, session four will be an introduction to the value of participatory design in ubicomp, session five will discuss the user experience implications of the physical and social contexts in scenarios of intelligent networked interfaces and the sixth session will focus on understanding how innovative ethno-methodologies and other advanced in-situ evaluation techniques can be useful for the identification of soft-data requirements (such as emotions and feelings) leading to the design and assessment of feasible and relevant future communications scenarios.

Instructors
Dr. Anxo Cereijo Roibs is Senior Lecturer at the University of Brighton, Visiting Lecturer at Westminster University, at the Politecnico di Milano and the National Institute of Design (India). His expertise resides in the user experience in pervasive communication systems. He has been HCI Manager at the Mobile Internet Services Provider, HiuGO SpA, and User Experience Consultant for Vodafone and since 2000, he has been collaborating with the Nokia Research Center. He has coordinated European funded projects such as the TRAIN-IT, European Union - Canada Program for Cooperation in Higher Education and Training. He has coordinated an ethnographic research addressing the future use of mobile phones as multimedia tools in collaboration with the Vodafone Group Foundation and the British Royal Academic of Engineering. He is British Telecom Fellow at the BT IT Mobility Research Centre and Executive Committee member of the British HCI Group.

Antonio Krueger is an associate professor for Geoinformatics and Computer Science at Mnster University, Germany and the managing director of the Institute for Geoinformatics (ifgi) at the same University. Antonio is also a co-founder of Eyeled GmbH, a company focusing on mobile computing solutions, such as the annual mobile fair guide for CeBIT. Antonio's main research areas include Intelligent User Interfaces and Mobile Context-aware Systems. He worked on the automatic generation of graphics for technical documentations, intelligent navigation systems and personalized media generation. In this context he looked at generation processes that take into account both the limited technical resources of output devices and the limited cognitive resources of the users. He co-organizes the annual Smart-Graphics Symposium and served on various program committees in the field of intelligent mobile systems and instrumented environments, e.g. the international conferences on Intelligent User Interfaces, User Modeling, Ubiquitous, and Pervasive Computing.

[top]



W1: (CANCELLED) Workshop on Multi-User and Ubiquitous User Interfaces (MU3I)

The Ubiquitous Computing paradigm has the potential of significantly changing the way in which users interact with computers by providing virtually ubiquitous access to services and applications through a large number of cooperating devices. However, in order to make this vision come true in a usable way, several challenges must be addressed. Supporting multiple users with this technology poses a challenge to several aspects of ubiquitous computing, such as the unpredictability of group setups and task selection when using everyday objects or environments with embedded technology. Even if embedded technologies are integrated with established social protocols, they may challenge these protocols, e.g. by introducing new means to seizing control of a shared resource. The workshop aims to be a platform where these issues will be discussed.

Organizer: Christian Kray, Newcastle Univ., UK, and Andreas Butz, LMU Munich, Germany

URL:http://www.mu3i.org/


W2: Tangible Play: Research and Design for Tangible and Tabletop Games (ACM Digital Library Link)

Many people of all ages play games, such as board games, PC games or console games. They like game play for a variety of reasons: as a pastime, as a personal challenge, to build skills, to interact with others, or simply for fun. Some gamers prefer board games over newer genres, because it allows them to socialize with other players face-to-face, or because the game play can be very improvisational as players rework the rules or weave stories around an unfolding game. Conversely, other gamers prefer the benefits of digital games on PCs or consoles. These include high quality 3D graphics, the adaptive nature of game engines (e.g. increasing levels of difficulty based on player experience) and an abundance of digital game content to explore and experience. With the increasing digitization of our everyday lives, the benefits of these separate worlds can be combined in the form of tangible games. For example, tangible games can be played on digital tabletops that provide both an embedded display and a computer to drive player interactions. Several people can thus sit around the table and play digital games together.

Organizers: Elise van den Hoven, Eindhoven Univ., Netherlands, and Ali Mazalek, Georgia Tech, USA

URL: http://www.lcc.gatech.edu/~mazalek/workshops/iui2007-tangibleplay/


W3: Common Sense and Intelligent User Interfaces (ACM Digital Library Link)

Ideally, computer interfaces will be able to interact with users at a higher level by understanding our goals, our problems, and the social procedures by which we live. In order for these intelligent computer interfaces to see the world from the perspective of their users, they must have access to a wealth of information about the world that human users take for granted. This information, which forms the basis of goal-directed computer interactions, is common sense knowledge. Common sense knowledge is non-expert and possessed by every person. Thus, volunteers make up a significant source of common sense knowledge being collected today, and intelligent interfaces help these contributors create robust and complete common sense databases. An interactive and intelligent environment can guide a contributor to add the information that would be most useful to the system and, with good design, can make the knowledge entry experience more rewarding for the contributor. In turn, this collected common sense knowledge helps enable a wide variety of interfaces to function better. We're interested in exploring both sides of this symbiotic relationship. How can common sense enable computers interfaces to better understand their human users? How can interfaces enable the elicitation of common sense knowledge?

Organizers: Catherine Havasi, Brandeis University, USA, and Henry Lieberman, MIT Media Lab, USA

URL: http://eurydice.cs.brandeis.edu/csiui/

[top]


Invited Speaker: Susan Dumais
Monday, January 29th, 9:00 am - 10:30 am

 

Information Retrieval in Context (ACM Digital Library Link)
Susan Dumais (Microsoft Research)

Abstract
Most information retrieval technologies are designed to facilitate information discovery. However, much knowledge work involves finding and re-using previously seen information in the context of ongoing work activities. An overview of techniques that people currently use to support re-access will be presented, and usage experiences with the "Stuff I've Seen" desktop search prototype that provides unified access to a wide range of heterogeneous information that a person has previously encountered (email, web pages, files, news, appointments) will be summarized. Key finding include the importance of time and people as retrieval cues, and the importance of metadata in supporting interactive retrieval. Alternative presentation techniques that leverage rich contextual cues such as timelines and memory landmarks are promising alternatives to long ranked lists of search results. Richer personalized information retrieval capabilities can also be supported using this infrastructure since it provides a very rich client-side representation of a user's interests and their evolution over time.

About Susan Dumais
Susan Dumais is a Senior Researcher in the Adaptive Systems and Interaction Group at Microsoft Research. She has been at Microsoft Research since 1997 and has published widely in the areas of human-computer interaction and information retrieval. Her current research focuses on personal information management, user modeling and personalization, novel interfaces for interactive retrieval, and implicit measures of user interest and activity. She has worked closely with several Microsoft groups (Windows Desktop Search, MSN Search, SharePoint Portal Server and Office Help) on search-related innovations. Prior to joining Microsoft Research, she was at Bellcore and Bell Labs for many years, where she worked on Latent Semantic Indexing (a statistical method for concept-based retrieval), combining search and navigation, individual differences, and organizational impacts of new technology.

Susan has published more than 170 articles in the fields of information science, human-computer interaction, and cognitive science, and holds several patents on novel retrieval algorithms and interfaces. She is Past-Chair of ACM's Special Interest Group in Information Retrieval (SIGIR), and served on the NRC Committee on Computing and Communications Research to Enable Better Use of Information Technology in Digital Government, and the NRC Board on Assessment of NIST Programs. She is on the editorial boards of ACM: Transactions on Information Systems, ACM: Transactions on Human Computer Interaction, Human Computer Interaction, Information Processing and Management, Information Retrieval, New Review of Hypermedia and Multimedia, and the Annual Review of Information Science and Technology, an associate editor for the first and second editions of the Handbook of Applied Cognition, and on program committees for several conferences. She was elected to the CHI Academy in 2004. Susan is an adjunct professor in the Information School at the University of Washington, and has been a visiting faculty member at Stevens Institute of Technology, New York University, and the University of Chicago.

Additional information is available at:
http://research.microsoft.com/~sdumais

[top]


Papers: Recommender Systems
Monday, January 29th, 11:00 am - 12:30 pm

 

Lies and Propaganda: Detecting Spam Users in Collaborative Filtering (ACM Digital Library Link)
Bhaskar Mehta (Fraunhofer IPSI)
Thomas Hofmann (Darmstadt University of Technology)
Peter Fankhauser (Fraunhofer IPSI)

Abstract
Collaborative Filtering systems are essentially social systems which base their recommendation on the judgment of a large number of people. However, like other social systems, they are also vulnerable to manipulation by malicious social elements. Lies and Propaganda may be spread by a malicious user who may have an interest in promoting an item, or downplaying the popularity of another one. By doing this systematically, with eithermultiple identities, or by involving more people, a few malicious user votes and profiles can be injected into a collaborative recommender system. This can significantly affect the robustness of a system or algorithm, as has been studied in recent work [5, 7]. While current detection algorithms are able to use certain characteristics of spam profiles to detect them, they suffer from low precision, and require a large amount of training data. In this work, we provide a simple unsupervised algorithm, which exploits statistical properties of effective spam profiles to provide a highly accurate and fast detection algorithm for detecting spam.

Hybrid Critiquing-Based Recommender Systems (ACM Digital Library Link)
Li Chen and Pearl Pu (Swiss Federal Institute of Technology in Lausanne)

Abstract
We propose a novel critiquing-based recommender interface, the hybrid critiquing interface that integrates the user self-motivated critiquing facility to compensate for the limitations of system-proposed critiques. The results from a user study show that the integration of such self-motivated critiquing support enables users to achieve a higher level of decision accuracy while consuming less cognitive effort. In addition, users expressed higher subjective opinions of the hybrid critiquing interface than the interface simply providing system-proposed critiques, and they would more likely return to it for future use.

SuggestBot: Using Intelligent Task Routing to Help People Find Work in Wikipedia (ACM Digital Library Link)
Dan Cosley (Cornell University)
Dan Frankowski, Loren Terveen, John Riedl (University of Minnesota)

Abstract
Member-maintained communities ask their users to perform tasks the community needs. From Slashdot to IMDb to Wikipedia, groups with diverse interests have created community-maintained artifacts of lasting value (CALV) that support the group's main purpose and provide value to others. Most such communities don't help members find work to do, or do so without regard to individual preferences, such as Slashdot assigning meta-moderation randomly. Yet social science theory suggests that reducing the cost and increasing the personal value of contribution will motivate members to participate more.

We present SuggestBot, software that performs intelligent task routing (matching people with tasks) in Wikipedia. SuggestBot uses broadly applicable strategies of text analysis, collaborative filtering, and link following to recommend tasks. SuggestBot's intelligent task routing increases the number of edits by roughly four times over suggesting random articles. Our contributions are: 1) demonstrating the value of intelligent task routing in a real deployment, 2) showing how to do intelligent task routing, and 3) sharing our experiences deploying a tool in Wikipedia, which offered both challenges and opportunities for research.

[top]


Papers: Social Software
Monday, January 29th, 2:30 pm - 4:00 pm

 

From Social Bookmarking to Social Summarization: An Experiment in Community-Based Summary Generation (ACM Digital Library Link)
Oisin Boydell and Barry Smyth (University College Dublin)

Abstract
We describe a novel document summarization technique that uses informational cues, such as social bookmarks or search queries, as the basis for summary construction by leveraging the snippet-generation capabilities of standard search engines. A comprehensive evaluation demonstrates how the social summarization technique can generate summaries that are of significantly higher quality that those produced by a number of leading alternatives.

Collecting Community Wisdom: Integrating Social Search & Social Navigation (ACM Digital Library Link)
Jill Freyne and Barry Smyth (University College Dublin)
Rosta Farzan and Peter Brusilovsky (University of Pittsburgh)

Abstract
The goal of the paper is to explain the integration of two social Web technologies social search and social navigation to demonstrate benefits on two levels. First, both technologies harvest and harness community wisdom and in an integrated system each of the search and navigation/browsing components benefit from the additional community wisdom when it comes to assisting users to locate relevant information. Second, integrating search and browsing facilitates the development of a unique interface that effectively blends search and browsing functionality as part of a seamless social information access service that allows users to effectively combine their search and browsing behaviors. In this paper we will argue that this integration provides significantly more than the simple sum of the parts.

Talk Amongst Yourselves: Inviting Users To Participate In Online Conversations (ACM Digital Library Link)
Franklin Harper, Dan Frankowski, Sara Drenner, Loren Terveen, John Riedl(University of Minnesota)
Yuqing Ren, Sara Kiesler, Robert Kraut (Carnegie Mellon University)

Abstract
Many small online communities would benefit from increased diversity or activity in their membership. Some communities run the risk of dying out due to lack of participation. Others struggle to achieve the critical mass necessary for diverse and engaging conversation. But what tools are available to these communities to increase participation? Our goal in this research was to spark contributions to the movielens.org discussion forum, where only 2% of the members write posts. We developed personalized invitations, messages designed to entice users to visit or contribute to the forum. In two field experiments, we ask (1) if personalized invitations increase activity in a discussion forum, (2) how the choice of algorithm for intelligently choosing content to emphasize in the invitation affects participation, and (3) how the suggestion made to the user affects their willingness to act. We find that invitations lead to increased participation, as measured by levels of reading and posting. More surprisingly, we find that invitations emphasizing the social nature of the discussion forum are effective, while invitations emphasizing non-social aspects of the discussion are less so.

[top]


Papers: User Modeling
Tuesday, January 30th, 9:00 am - 10:30 am

 

Unsupervised and Supervised Machine Learning in User Modeling for Intelligent Learning Environments (ACM Digital Library Link)
Saleema Amershi and Cristina Conati (Department of Computer Science, University of British Columbia)

Abstract
We outline a user modeling framework that uses both unsupervised and supervised machine learning in order to reduce the costs of building user models and facilitate transferability. We apply the framework to model student learning during interaction with the Adaptive Coach for Exploration (ACE) learning environment (using both interface and eye-tracking data). In addition to demonstrating framework effectiveness, we also compare results from previous research on applying the framework to a different learning environment and data type. Our results also confirm previous findings on the value of using eye-tracking data to assess student learning.

Toward Harnessing User Feedback For Machine Learning (ACM Digital Library Link)
Simone Stumpf, Vidya Rajaram, Lida Li, Margaret Burnett, Thomas Dietterich, Erin Sullivan, Russell Drummond, Jonathan Herlocker (Oregon State University)

Abstract
There has been little research into how end users might be able to communicate advice to machine learning systems. If this resourcethe users themselvescould somehow work hand-in-hand with machine learning systems, the accuracy of learning systems could be improved and the users understanding and trust of the system could improve as well. We conducted a think-aloud study to see how willing users were to provide feedback and to understand what kinds of feedback users could give. Users were shown explanations of machine learning predictions and asked to provide feedback to improve the predictions. We found that users had no difficulty providing generous amounts of feedback. The kinds of feedback ranged from suggestions for reweighting of features to proposals for new features, feature combinations, relational features, and wholesale changes to the learning algorithm. The results show that user feedback has the potential to significantly improve machine learning systems, but that learning algorithms need to be extended in several ways to be able to assimilate this feedback.

Supporting Interface Customization Using a Mixed-Initiative Approach (ACM Digital Library Link)
Andrea Bunt, Cristina Conati, Joanna McGrenere (University of British Columbia)

Abstract
We describe a mixed-initiative framework designed to sup-port the customization of complex graphical user interfaces. The framework uses an innovative form of online GOMS analysis to provide the user with tailored customization sug-gestions aimed at maximizing the users performance with the interface. The suggestions are presented non-intrusively, minimizing disruption and allowing the user to maintain full control. The framework has been applied to a general user-productivity application. A formal user evaluation of the system provides encouraging evidence that this mixed-initiative approach is preferred to a purely adaptable alterna-tive and that the systems suggestions help improve task per-formance.

[top]


Papers: Information Retrieval
Tuesday, January 30th, 11:00 am - 12:30 pm

 

What Do People Recall About Their Documents? Implications for Desktop Search Tools (ACM Digital Library Link)
Tristan Blanc-Brude and Dominique L. Scapin (INRIA Rocquencourt)

Abstract
This study aims at finding out which attributes people actually recall about their own documents (electronic and paper), and what are the characteristics of their recall, in order to provide recommendations on how to improve tools allowing users to retrieve their electronic files more effec-tively and more easily. An experiment was conducted with fourteen participants at their workplace. They were asked first to recall features about one (or several) of their own work documents, and secondly to retrieve these docu-ments. The difficulties encountered by the participants in retrieving their electronic documents support the need for better retrieval tools. More specifically, results of the recall task indicate which attributes are candidates for facilitating file retrieval and how search tools should use these attributes.

Mobile Content Enrichment (ACM Digital Library Link)
Karen Church and Barry Smyth (University College Dublin)

Abstract
Delivering an effective mobile search service is challenging for many reasons. Certainly small-screen mobile handsets with limited text input capabilities do not make ideal search devices. In addition, the brevity of mobile Internet content hampers effective indexing and limits retrieval opportunities. In this paper we focus on this indexing issue and describe an approach that leverages Web search engines as a source of content enrichment. We present an evaluation using a mobile news service that demonstrated significant improvements in search performance compared to a standard benchmark system.

Context-Aware Adaptive Information Retrieval for Investigative Tasks (ACM Digital Library Link)
Zhen Wen, Michelle Zhou, Vikram Aggarwal (IBM T. J. Watson Research Center)

Abstract
We are building an intelligent information system to aid users in their investigative tasks, such as detecting fraud. In such a task, users must progressively search and analyze relevant information before drawing a conclusion. In this paper, we address how to help users find relevant information during an investigation. Specifically, we present a novel approach that can improve information retrieval by exploiting a users investigative context. Compared to existing retrieval systems, which either are context insensitive or leverage only limited user context, our work offers two unique contributions. First, our system works with users cooperatively to build an investigative context, which is otherwise very difficult to achieve by machine or human alone. Second, we develop a context-aware method that can adaptively retrieve and evaluate information relevant to an ongoing investigation. Experiments show that our approach can improve the relevance of retrieved information significantly. As a result, users can fulfill their investigative tasks more efficiently and effectively.

[top]


Papers: Personal Assistants
Tuesday, January 30th, 2:30 pm - 4:00 pm

 

Active EM to Reduce Noise in Activity Recognition (ACM Digital Library Link)
Jianqiang Shen and Thomas Dietterich (School of EECS, Oregon State University)

Abstract
Intelligent desktop environments allow the desktop user to define a set of projects or activities that characterize the user's desktop work. These environments then attempt to identify the current activity of the user in order to provide various kinds of assistance. These systems take a hybrid approach in which they allow the user to declare their current activity but they also employ learned classifiers to predict the current activity to cover those cases where the user forgets to declare the current activity. The classifiers must be trained on the very noisy data obtained from the user's activity declarations. Instead of asking the user to review and relabel the data manually, we employ an active EM algorithm that combines the EM algorithm and active learning. EM can be viewed as retraining on its own predictions. To make it more robust, we only retrain on those predictions that are made with high confidence. For active learning, we make a small number of queries to the user based on the most uncertain instances. Experimental results on real users show this active EM algorithm can significantly improve the prediction precision, and that it performs better than either EM or active learning alone.

Entropy-Driven Online Active Learning for Interactive Calendar Management (ACM Digital Library Link)
Julie Weber and Martha Pollack (University of Michigan)

Abstract
We present a new algorithm for active learning embedded within an interactive calendar management system that learns its users' scheduling preferences. When the system receives a meeting request, the active learner selects a set of alternative solutions to present to the user; learning is then achieved by noting the user's preferences for the selected schedule over the others presented. This application imposes a number of constraints on the active learning process. Most notably, and distinct from many active learning settings, arbitrary training examples cannot be generated; instead, learning opportunities only arise when the system receives a request for a new meeting, and the only possibilities for presentation are valid schedules that incorporate the requested meeting. In addition, because learning occurs during actual system use, our active learning component must select schedules for presentation that meet the user's needs as well as enhance the learning process. To achieve these goals, we introduce a new approach to active learning that makes online decisions about the technique to use in selecting the schedules to present in response to a meeting request. The decision is based on the entropy of the available options. We present experimental results that indicate that our entropy-driven active learning algorithm allows us to balance learning efficiency and user satisfaction better than static selection techniques.

Segmenting Meetings into Agenda Items by Extracting Implicit Supervision from Human Note-Taking (ACM Digital Library Link)
Satanjeev Banerjee and Alexander I. Rudnicky (Carnegie Mellon University)

Abstract
Splitting a meeting into segments such that each segment contains discussions on exactly one agenda item is useful for a lot of tasks, such as retrieval and summarization of agenda item discussions. However, topic segmentation of meetings is a difficult task. In this paper, we investigate the idea of acquring implicit supervision from the human meeting participants to solve the segmentation problem. Specifically we have implemented and tested a note taking interface that gives value to the user by helping him organize and retrieve his notes easily, but that extracts a segmentation of the meeting simply based on his note taking behavior. We show that the segmentation so obtained achieves a Pk value of 0.163 which improves upon an unsupervised baseline by 56%, and also compares favorably with current published results. Most importantly, we achieve this performance without any "features" or "algorithms" in the classic sense.

[top]


Papers: Demonstration Based Interfaces
Tuesday, January 30th, 4:30 pm - 5:30 pm

 

Distributed Augmentation-Based Learning, a Learning Algorithm for Distributed Collaborative Programming-by-Demonstration (ACM Digital Library Link)
Vittorio Castelli, Lawrence Bergman (IBM T.J. Watson Research Center)

Abstract
The learning algorithms used in Programming-by-Demonstration (PBD) are either on-line and incremental or off-line and batch. Neither category is entirely suitable for capturing know-how from demonstrations in a distributed, collaborative environment, where multiple expert can independently provide examples to improve the model.

In this paper we describe Distributed Augmentation-Based Learning (DABL), the first real-time PBD learning algorithm suited for distributed know-how acquisition. DABL is an incremental learning algorithm that uses a version-control-like paradigm to combine independently constructed procedure models. An expert can check out a procedure model from a repository and modify it by means of new demonstrations or by manually editing it. The expert then reconciles the changes with those concurrently made by other experts and checked into the repository.

DABL automatically merges the two procedures, learns new decision points based on reconcilable differences, and identifies conflicts where there are multiple valid ways of combining the changes or where the combination produces an invalid model, that is, one that does not lie in the search space of the learning algorithm.

Building Data Integration Queries by Demonstration (ACM Digital Library Link)
Rattapoom Tuchinda, Craig Knoblock, Pedro Szekely (University of Southern California)

Abstract
The magnitude of data available on the web prompts the need for an easy to use query interface that enables the users to integrate data from multiple web sources in an intelligent fashion. Past work in the area of databases has resulted in different query interface systems that simplify query formulation. While these approaches reduce the users effort to compose queries, the user is still required to pick data sources to use and the interaction does not guarantee to yield a non-empty result set. We introduce a novel approach that exploits the structure of the relational data source(s) to formulate a set of constraints. These constraints are used in conjunction with partial plans to produce an intelligent query interface that (a) does not require the user to know details about data sources or existing values (b) suggests valid inputs to the user (c) creates consistent query that always return values.

[top]


Poster Session
Tuesday, January 30th, 7:00 pm - 9:00 pm

 

iMime: An Interactive Character Animation System for use in Dementia Care (ACM Digital Library Link)
Andreas Wiratanaya, Michael Lyons, Shinji Abe (ATR Intelligent Robotics and Communication Labs)
Nicholas Butko (Department of Cognitive Science, UC San Diego)

Abstract
We describe the design and implementation of an interactive character animation interface. The system analyzes the attentive state and aspects of the affective behaviour of a viewer using input from a video camera and uses this to control the behaviour of a cartoon-like animated character. Using the interaction metaphor of a mime artist, we design the system to encourage viewer attention and interaction, with adaptation using an online reinforcement learning based on the viewers attentive state. This work is ultimately aimed at developing a system to support the care of dementia sufferers.

Robotic Telecommunication System based on Facial Information Measurement (ACM Digital Library Link)
Junichi Ido, Etsuko Ueda, Yoshio Matsumoto, Tsukasa Ogasawara (Nara Institute of Science and Technology)

Abstract
This paper proposes a multi-modal telecommunication system using a facial expression robot. We developed a telecommunication system which projects facial expression of an operator to a remote place using facial expression robot "Infanoid2." The facial information of a user is measured using a stereo camera system and is projected to a robot used to communicate with another person in a remote location. Experiment of the impression evaluation is performed using this system. This paper discusses the effectiveness of robots as a telecommunication medium based on the experimental results.

A Data-Oriented Approach to Integrate Emotions in Adaptive Dialogue Management (ACM Digital Library Link)
Johannes Pittermann and Angela Pittermann (University of Ulm)

Abstract
The involvement of emotions in dialogue design has attracted much interest in current research on human-computer interfaces during the past years. We focus on the implementation of a flexible and robust dialogue system which is able to include emotions and other influencing parameters in the dialogue flow. As opposed to widely-used rule-based systems, which are rather unflexible in terms of portability and adaptability, we propose a simplified stochastic approach to model dialogue management on the basis of user feedback.

MathBox: Interactive Pen-based Interface for Inputting Mathematical Expressions (ACM Digital Library Link)
Yuji Kasuya and Hayato Yamana (Waseda University)

Abstract
Inputting mathematical expressions with a mouse and a keyboard is a troublesome task. Thus, a number of mathematical expression recognition systems capable of recognizing handwritten mathematical expressions to input them into computers have been proposed. Even with these systems, however, structure recognition of mathematical expressions is still difficult. This paper proposes a new pen-based interface for inputting mathematical expressions into computers. The proposed interface interactively shows "boxes" in which the user can write one symbol. The boxes are shown along with the user's writing. For example, when the user writes 'x,' the interface shows the boxes for a power and an index of 'x' and for the next symbol. When the user inputs a fraction line, the interface shows boxes for the numerator, denominator, and the next symbol. The proposed interface skips recognizing the structures of expressions, which enables users to write mathematical expressions with practical accuracy.

Analysis of Affect Expressed through the Evolving Language of Online Communication (ACM Digital Library Link)
Alena Neviarouskaya and Mitsuru Ishizuka (University of Tokyo)
Helmut Prendinger (National Institute of Informatics)

Abstract
In this paper, we focus on affect recognition from text in order to facilitate sensitive and expressive communication in computer-mediated environments. Our model for analyzing affect conveyed by text is tailored to handle the style and specifics of informal online conversations. The motivation behind our approach is to improve social interactivity and emotional expressiveness of real-time messaging.

In order to estimate affect in text, our model processes symbolic cues, such as emoticons, detects and transforms abbreviations, and employs natural language processing techniques for word/phrase/sentence-level analysis, e.g. by considering relations among words in a sentence. As a result of the analysis, text can be categorized into emotional states and communicative functions. A designed graphical repre-sentation of a user (avatar) displays emotions and social behaviour driven by text and performs natural idle move-ments. The proposed system shows promising results on affect recognition in real examples of online conversation.

On the Community-Based Explanation of Search Results (ACM Digital Library Link)
Maurice Coyle and Barry Smyth (University College Dublin)

Abstract
Collaborative Web search (CWS) is an approach to personalizing search results, returned by an underlying search engine(s), to the preferences of a community of like-minded searchers. In this paper we propose an alternative architecture that facilitates a more flexible integration between CWS and the underlying search engine(s) and evaluate how community behaviour can be used to annotate search results with explanatory information to facilitate relevancy judgments.

Social Radio A Music-Based Approach to Emotional Awareness Mediation (ACM Digital Library Link)
Richard Etter, Carsten Rcker (Fraunhofer IPSI)

Abstract
This paper presents a novel approach for mediating awareness in small intimate groups. Instead of traditional communication media, music is used to inform users about the presence and mood of multiple remote peers. Based on this conceptual idea, an awareness system called 'Social Radio' was developed. The system consists of several smart artifacts and an underlying multi-user communication infrastructure.

ONCOR: Ontology- and Evidence-based Context Reasoner (ACM Digital Library Link)
Judy Kay, William Niu, David Carmichael (University of Sydney)

Abstract
In this paper, we describe ONCOR, our ontology- and evidence-based approach to reason about context. ONCOR tackles the critical problem of providing a flexible and pragmatic approach to building lightweight ontologies of places, devices and sensors in MyPlace, a ubiquitous computing application which provides personalised information about a building. Our contributions lie in constructing a general middle ontology for a building (MIBO) based on OpenCyc, combined with our own approach to evidence-based reasoning, (accretion and resolution). We describe the evaluation of ONCOR in terms of a comparison of its answers to core ontological questions about context with previous work.

Refining Preference-Based Search Results Through Bayesian Filtering (ACM Digital Library Link)
Jiyong Zhang and Pearl Pu (Swiss Federal Institute of Technology)

Abstract
Preference-based search(PBS) is a popular approach for helping consumers find their desired items from online catalogs. Currently most PBS tools generate search results by a certain set of criteria based on preferences elicited from the current user during the interaction session. Due to the incompleteness and uncertainty of the user's preferences, the search results are often inaccurate and may contains items that the user has no desire to select. In this paper we develop an efficient Bayesian filter based on a group of users' past choice behavior and use it to refine the search results by filtering out items which are unlikely to be selected by the user. Our preliminary experiment shows that our approach is highly promising in generating more accurate search results and saving user's interaction effort.

Personalized Ambient Media Experience: move.me Case Study (ACM Digital Library Link)
Lora Aroyo and Hielke Schut (Eindhoven University of Technology)
Frank Nack, Thecla Schiphorst, Michiel KauwATjoe (V2_ Institute for Instable Media)

Abstract
The move.me prototype illustrates a scenario for social interaction in which users can manipulate audio-visual sources presented on various screens through an interaction with a sensor-enhanced pillow. The technology developed for move.me uses the surface of a pillow as a tactile interface. We describe the underlying concepts of move.me and its motivations. We present a case study of the environment as the context of evaluating aspects of our approach and conclude with plans for future work.

Auditory perceptible landmarks in mobile navigation (ACM Digital Library Link)
Joerg Baus, Rainer Wasinger, Ilhan Aslan (DFKI GmbH)
Antonio Krger (University of Mnster)
Andreas Maier and Tim Schwartz (Saarland University)

Abstract
Normally, mobile pedestrian navigation systems use visually perceptible landmarks to guide their users through the environment. In this article we introduce concepts for the use of auditory perceptible landmarks in route descriptions. Such auditory perceptible landmarks complement their visual counterparts and also stand to be beneficial for certain groups like the visually impaired and the elderly.

Supporting Small Groups in the Museum by Context-Aware Communication Services (ACM Digital Library Link)
Tsvi Kuflik, Julia Sheidin, Sadek Jbara, Dina Goren-Bar, Pnina Soffer (The University of Haifa)
Oliviero Stock and Massimo Zancanaro (ITC/irst)

Abstract
Visitors often tend to visit museums in groups, mainly with family or friends, yet most of the today mobile museum guides focus on supporting the individual visitor. The technology described in this paper allows supporting groups of visitors in addition to individuals by providing context-aware services aimed at supporting the whole group. These include context-aware communication and alerting services that are provided by the museum visitor's guide system developed in the framework of the PIL (PEACH-Israel) project, as an example case of a larger variety of possible context-aware services.

Towards Intelligent Mapping Applications: A Study Of Elements Found In Cognitive Maps (ACM Digital Library Link)
Gary Look and Howard Shrobe (MIT Computer Science and Artificial Intelligence Laboratory)

Abstract
This paper describes a study that examines common elements found in people's mental maps of the city of Boston. The intent of the study is to understand the mental model people have of a city. Understanding this mental model will provide insight into developing mapping applications that present location information in a way that makes it easier to conceptualize and situate new location information in terms of places a person already knows. An analysis of hand-annotated maps showing the locations of prominent places in a person's mental map of Boston suggests that prominent places can be characterized by a certain set of properties and that major transit points (subway stops) play an important role in framing a person's mental map.

Modelling Personality in Voices of Talking Products Through Prosodic Parameters (ACM Digital Library Link)
Michael Schmitz (DFKI GmbH)
Antonio Krger (University of Mnster)
Sarah Schmidt (Saarland University)

Abstract
In this paper we report preliminary findings from two user studies that on the one hand investigate how prosodic parameters of synthetic speech can influence the perceived impression of the speakers personality and on the other hand explores if and how people attribute personality to objects such as typical products of daily shopping. The results show that a) prosodic parameters have a strong influence on the perceived personality and can be partially used to achieve a desired impression and b) that subjects clearly attribute personalities to products. Both findings encourage us to continue our work on a dialogue shell for talking products.

A Comparison of Two Compound Critiquing System (ACM Digital Library Link)
James Reilly, Lorraine McGinty, Barry Smyth (Adaptive Information Cluster, School of Computer Science & Informatics)
Jiyong Zhang, Pearl Pu (Human Computer Interaction Group, Swiss Federal Institute of Technology)

Abstract
Compound critiques allow users to simultaneously express directional preferences over several product attributes. Presenting the user with compound critiques is not a new idea. The original Find-Me Systems (e.g., Car Navigator) showed static compound critiques; they didn't change irrespective of user preferences or the product availability. Recently, a number of techniques for dynamically generating compound critiques have been proposed. While these techniques have been evaluated in isolation, to date no direct comparison of these (in terms of their interfacing characteristics and recommendation performance) has been reported. Motivated by this, our research groups have came together to carry out this comparison for the approaches we each take. The user study platform that we have developed facilitates the comparison of various critiquing based recommenders. In this paper we report the first set of results from a comprehensive live-user evaluation of two dynamic compound critique systems using this evaluation platform.

User-Context for Adaptive User Interfaces (ACM Digital Library Link)
Anil Shankar, Sushil J. Louis, Sergiu Dascalu, Linda J. Hayes, Ramona Houmanfar (University of Nevada)

Abstract
We present results from an empirical user-study with ten users which investigates if information from a user's environment helps a user interface to personalize itself to individual users to better meet usability goals and improve user-experience. In our research we use a microphone and a web-camera to collect this information (user-context) from the vicinity of a subject's desktop computer. Sycophant, our context-aware calendaring application and research test-bed uses machine learning techniques to successfully predict a user-preferred alarm type. Discounting user identity and motion information significantly degrades Sycophant's performance on the alarm prediction task. Our user study emphasizes the need for user-context for personalizable user interfaces which can better meet effectiveness and utility usability goals. Results from our study further demonstrate that contextual information helps adaptive interfaces to improve user-experience.

Exploiting Web Browsing Histories to Identify User Needs (ACM Digital Library Link)
Fabio Gasparetti and Alessandro Micarelli (University of Roma Tre)

Abstract
Browsing activities are an important source of information to build profiles of the user interests and personalize the humancomputer interaction during information seeking tasks. Visited pages are easily collectible, e.g., from browsers histories and toolbars, or desktop search tools, and they often contain documents related to the current user needs. Nevertheless, menus, advertisements or pages that cover multi topics affect negatively the advantages of an implicit feedback technique that exploits these data to build and keep updated user profiles. This work describes a technique to collect text relevant to the current needs from sequences of pages visited by the user. An evaluation shows how it outperforms other techniques that consider the whole page contents. We also introduce an improvement based on machine learning techniques that is currently under evaluation.

Emotionally Reactive Television (ACM Digital Library Link)
Chia-Hsun Lee, Chaochi Chang, Hyemin Chung, Ted Selker (MIT Media Lab)

Abstract
When is an interface simple? Is it when it is invisible or very obvious, even intrusive? From the time TV was created, watching TV is considered as a static activity. TV audiences have very limited choices to interact with TV, such as turning on/off, increasing/decreasing volume, and traversing among different channels. This paper suggests that TV program should have social responses to people, such as affording and accepting audience's emotional feeling with the growth of technologies. This paper presents HiTV, an Emotionally-Reactive TV system using a digitally augmented soft ball as affect-input interfaces that can amplify TV program's video/audio signals. HiTV transforms the original video and audio into effects that intrigue and fulfill people's emotional expectation.

A Markup Language for Describing Interactive Humanoid Robot Presentations (ACM Digital Library Link)
Yoshitaka Nishimura, Shinichiro Minotsu, Hiroshi Dohi, Mitsuru Ishizuka (University of Tokyo)
Mikio Nakano, Kotaro Funakoshi, Johane Takeuchi, Yuji Hasegawa, Hiroshi Tsujino (Honda Research Institute Japan Co., Ltd.)

Abstract
This paper presents a multi-modal presentation markup language for humanoid roboalgorithmts, MPML-HR ver 3.0, which is able to describe presentation contents including speech-based interactions with audiences. Previous versions of MPML-HR does not feature any interaction functionality which dynamically changes the presentation according to the utterances by audiences, although such interaction makes the presentation more effective and understandable. Since MPML-HR ver 3.0 inherits simple descriptions of previous versions of MPML-HR, the content designer can describe interactive presentation without configuring conventional complicated multi-modal interaction systems.

Trusted Search Communities (ACM Digital Library Link)
Peter Briggs and Barry Smyth (University College Dublin)

Abstract
We describe a social search technique that harnesses the search experiences of a community of searchers to generate result recommendations, in a collaborative fashion, to complement results returned from some underlying search engine. We describe a dynamic model of trust for this search network and provide experimental results to show that search performance improves as the network evolves.

Combating Information Overload in Non-Visual Web Access Using Context (ACM Digital Library Link)
Jalal Mahmud, Yevgen Borodin, Dipanjan Das, I.V. Ramakrishnan (Stony Brook University)

Abstract
Web sites are designed for graphical mode of interaction. Sighted users can visually segment Web pages and quickly identify relevant information. On the contrary, individuals with visual disabilities have to use screen readers to browse the Web. As screen readers process pages sequentially and read through everything, Web browsing becomes time-consuming and strenuous. Although, the use of shortcut keys and searching offers some improvements, the problem still remains. In this paper, we address this problem using the notion of context. Our prototype system, CSurf, embodying our approach, provides all features of a usual screen reader. However, when a user follows a link, CSurf captures the context of the link, processes it with several NLP techniques, and uses it to identify relevant information on the next page. CSurf rearranges the content of the page, so that the relevant information is read out first. We conducted a series experiments to evaluate the performance of CSurf against the state-of-the-art screen reader JAWS. Our results show that the use of context can save browsing time and substantially improve the browsing experience of visually disabled individuals.

Temporal filtering system for reducing the risk of spoiling a user's enjoyment (ACM Digital Library Link)
Satoshi Nakamura, Katsumi Tanaka (Kyoto University)

Abstract
There are many methods available for filtering undesirable Web content. However, they have so far only focused on full-time filtering, which blocks harmful contents such as adult websites and viruses. In daily-content browsing, people need not only full-time filtering but also temporal filtering. For example, users do not want to see the final score of a sports match before watching it on TV. On the other hand, he/she might like to see the results after watching it. Our filtering system detects what types of content a user would find undesirable by monitoring the user's schedules and mo-tivations and then removing or hiding that content. In this paper, we propose a temporal filtering system and describe its implementation.

Increasing Web Accessibility by Automatically Judging Alternative Text Quality (ACM Digital Library Link)
Jeffrey Bigham (University of Washington)

Abstract
The lack of appropriate alternative text for web images remains a problem for blind users and others accessing the web with non-visual interfaces. The content contained within web images is vital for understanding many web sites but the majority are assigned either inaccurate alternative text or none at all. The capability to automatically judge the quality of alternative text has the promise to dramatically improve the accessibility of the web by bringing intelligence to three categories of interfaces: tools that help web authors verify that they have provided adequate alternative text for web images, systems that automatically produce and insert alternative text for web images, and screen reading software. In this paper we describe a classifier with the capability of measuring the quality of alternative text given only a few labeled training examples by automatically considering the image context.

Social Robots as Mediators between Users and Smart Environments (ACM Digital Library Link)
Berardina De Carolis and Giovanni Cozzolongo (University of Bari)

Abstract
In this paper we propose the use of a social robot as mediator between the user and services of a smart environment. We focus, in particular, on the need for the robot to comprehend the user's intention in order to respond accordingly. Since the speech is considered one of the more natural and immediate input channel in human-robot interaction we discuss the importance of recognizing, besides the linguistic content of the spoken sentence, the valence of the user tone of voice in order to infer properly the user's communicative intention during the interaction.

Modeling User Behavior Using a Search-Engine (ACM Digital Library Link)
Maeve O'Brien and Mark T. Keane (University College Dublin)

Abstract
A model of user-search-engine interaction within the ACT-R cognitive architecture is developed using rational analysis to characterize optimal behavioral strategies for information search using a search engine. We demonstrate, using an em-pirical evaluation, how the model, at an individual level, ties to studies describing the aggregate behaviors of large num-bers of users searching the Web. These results are discussed in terms of their practical implications for search interfaces and ranking algorithms.

Interactive Visual Clustering (ACM Digital Library Link)
Marie desJardins, James MacGlashan (MAPLE Laboratory)
Julia Ferraioli (Bryn Mawr College)

Abstract
Interactive Visual Clustering (IVC) is a novel method that allows a user to explore relational data sets interactively, in order to produce a clustering that satisfies their objectives. IVC combines spring-embedded graph layout techniques with user interaction and constrained clustering. This paper describes the IVC method, and gives experimental results on several synthetic and real-world data sets, showing that IVC yields better clustering performance than several alternative methods.

What am I gonna wear?: Scenario-Oriented Recommendation (ACM Digital Library Link)
Edward Shen, Henry Lieberman, Francis Lam (MIT Media Laboratory)

Abstract
Electronic Commerce on the Web is thriving, but consumers still have trouble finding products that will meet their needs and desires. AI has offered many kinds of Recommender Systems, but they are all oriented toward searching based on concrete attributes of the product (e.g. price, color) or the user (as in Collaborative Filtering). We introduce a novel technique, Scenario-Oriented Recommendation, which works even when users don't necessarily know exactly what product characteristics they are looking for.

Users describe a goal for a real-life scenario in which the desired product might be used, e.g. "I want something elegant to wear for my boss's birthday party". We use a Commonsense reasoning system to map between the goals stated by the user, and possible characteristics of the product that might be relevant. For example, the boss's birthday party suggests a higher value for the "formal vs. casual" attribute, than say, a child's birthday party. Reasoning is based on an 800,000-sentence Common Sense knowledge base, and spreading activation inference. Scenario-oriented recommendation breaks down boundaries between products' categories, finds the "first example" for existing techniques like Collaborative Filtering, and helps promote independent brands. We describe our scenario-oriented fashion recommendation system, What Am I Gonna Wear?.

VizScript: A High-Level Language for Rapidly Creating Custom Visualizations for Multi-Agent Systems (ACM Digital Library Link)
Jing Jin, Rajiv Maheswaran, Romeo Sanchez, Pedro Szekely (Information Sciences Institute, University of Southern California)

Abstract
We address the problem of creating visualizations to debug and understand multi-agent systems. The key challenges are that (1) needs arise dynamically, i.e., it is difficult to know a priori what visualizations one wants, (2) extensive expertise on the system, the algorithms and visualization tools are often needed for implementation, and (3) agents can be running in a distributed environment. We have developed VizScript, a collection of tools that expedites the process of creation. VizScript does not require great knowledge of the intricacies of internal data structures or display creation and is functional with data streams from multiple sources. We show that by combining generic instrumentation, a knowledge base, and simple scene definition primitives with a reasoning system, we are able to recreate the visualizations for a complex multi-agent system with an order-of-magnitude less effort than was required in a Java implementation.

[top]


Invited Speaker: Thomas Strothotte
Wednesday, January 31st, 9:00 am - 10:30 am

 

Image-Text Interaction (ACM Digital Library Link)
Thomas Strothotte (University of Rostock)

Abstract
In most interactive systems, the handling of text and images are still quite separate matters. Rendering pipelines handle 2D and 3D graphics, while text tends to be an after-thought which is hardly integrated at the system level, let alone at the level of user interaction.

This talk will survey recent developments in the area of image-text interaction from both the system architecture point of view and with respect to user interfaces. The talk will emphasize interactive labelling of 2D and 3D graphics and outline new real-time algorithms for label placement. Labelled images will be discussed as an interface to methods and tools for data mining. Finally, challenges for future developments in the area of image-text interaction will be outlined.

About Thomas Strothotte
Thomas Strothotte, University of Rostock, Germany, is a professor of computer science specializing in computer graphics and interactive systems. He was educated in Canada (BSc 1980, MSc 1981 Simon Fraser University, Vancouver; PhD 1984 McGill University) and has had previous academic appoints at various institutions including INRIA Rocquencourt, the IBM Scientific Center in Heidelberg, the Free Univ. of Berlin and the Univ. of Magdeburg, Germany. He has published extensively in the area of non-photorealistic computer graphics and human-computer interaction. Dr. Strothotte has held numerous administrative positions, and since October, 2006 he has been the President & CEO of the University of Rostock.

[top]


Papers: Natural Language Interfaces
Wednesday, January 31st, 11:00 am - 12:30 pm

 

Porting Natural Language Interfaces Between Domains: An Experimental User Study with the ORAKEL System (ACM Digital Library Link)
Philipp Cimiano, Peter Haase, Jrg Heizmann (Institute AIFB, University of Karlsruhe)

Abstract
We present a user-centered model for porting natural language interfaces (NLIs) between domains efficiently. The model assumes that domain experts without any background knowledge about computational linguistics will perform the customization of the NLI to a specific domain. In fact, it merely requires familiarity with the underlying knowledge base as well as with a few basic subcategorization types. Our model is iterative in the sense that the adaption of the NLI is performed in several cycles on the basis of the questions which the NLI failed to answer, thus iteratively increasing the coverage of the system. We provide experimental evidence in form of a user study as well as a case study involving a real-world application corroborating that our model is indeed a feasible way of customizing the interface to a certain domain.

BuzzTrack: Topic Detection and Tracking in Email (ACM Digital Library Link)
Gabor Cselle, Keno Albrecht, Roger Wattenhofer (ETH Zurich)

Abstract
We present BuzzTrack, an email client extension that helps users deal with email overload. This plugin enhances the interface to present messages grouped by topic, instead of the traditional approach of organizing email in folders. We discuss a clustering algorithm that creates the topic-based grouping, and a heuristic for labeling the resulting clusters to summarize their contents. Lastly, we evaluate the clustering scheme in the context of existing work on topic detection and tracking (TDT) for news articles: Our algorithm exhibits similar performance on emails as current work on news text. We believe that BuzzTrack's organization structure, which can be obtained at no cost to the end user, will be helpful for managing the massive amounts of email that land in the inbox every day.

Knowledge Acquisition from Simplified Text (ACM Digital Library Link)
Kevin Livingston and Christopher Riesbeck (Northwestern University)

Abstract
The problem of entering and integrating new knowledge into a logic-based knowledge base is substantial. Our solution is to provide a natural language interface, which reads simplified English, enabled by a knowledge-based memory-retrieval driven natural language understander. This paper presents a set of tools and interfaces for interacting with such a system, and a discussion of the underlying Reader system, the reading component of the Learning Reader project. The interfaces presented provide direct feedback about what portions of the text are understood, and what interpretations are being produced from it. In addition tools are presented to, among other things, provide example sentences, to facilitate users producing simplified English text suitable for the Reader.

[top]


Papers: Multi-Modal Interfaces
Wednesday, January 31st, 2:30 pm - 4:00 pm

 

A Platform for Output Dialogic Strategies in Natural Multimodal Dialogue Systems (ACM Digital Library Link)
Meriam Horchani (France Tlcom R&D, CLIPS-IMAG)
Laurence Nigay (CLIPS-IMAG)
Franck Panaget (France Tlcom R&D)

Abstract
The development of natural multimodal dialogue systems remains a very difficult task. The flexibility and naturalness they offer result in an increased complexity that current software tools do not address appropriately. One challenging issue we address here is the generation of cooperative responses in an appropriate multimodal form, highlighting the intertwined relation of content and presentation. We identify a key component, the dialogic strategy component, as a mediator between the natural dialogue management and the multimodal presentation. Among potential semantic information to present, this component selects the content to be presented according to various presentation constraints. Constraints include inherent characteristics of modalities, the availability of a modality as well as preferences of the user. Thus the cooperative behaviour of the system could be adapted as well as its multimodal behaviour. In this paper, we present the dialogic strategies component and an associated platform to quickly develop output multimodal cooperative responses in order to explore different dialogic strategies.

Music Compositional Intelligence with an Affective Flavor (ACM Digital Library Link)
Roberto Legaspi, Yuya Hashimoto, Koichi Moriyama, Satoshi Kurihara, Masayuki Numao (Institute of Scientific and Industrial Research, Osaka University)

Abstract
The consideration of human feelings in automated music generation by intelligent music systems, albeit a compelling theme, has received very little attention. This paper reports the performance outcome of a system that induces previously unknown coupling of musical structures and listenerfs impressions of various musical pieces and subsequently uses this induced relation as a personalized model to compose new tunes that are supposed to be perceived by the listener as tailored to his preference. From a consolidated set of listenerfs evaluations of musical pieces, the system employs multiple-part learning, which integrates inductive logic programming and multiple-instance learning, to induce a first-order rule modeling of the music structures-impressions relations. Subsequently, the system acquires music frame structures and chord progressions using a genetic algorithm whose fitness function is based upon the acquired model. Finally, the system creates melody out of the obtained chord and frame structures. The empirical results show that the system can acquire significant relations that can aid the adaptive improvisation of user-specific tunes amid the absence of expertise in music composition.

Making Sense of Virtual Environments: Action Representation of Grounding and Common Sense (ACM Digital Library Link)
Jean-Luc Lugrin and Marc Cavazza (School of Computing, University of Teesside)

Abstract
The development of complex interactive 3D systems raises the need for representations supporting more abstract descriptions of world objects, their behaviour and the world dynamics. The inclusion of Artificial Intelligence representations and their use within 3D graphic worlds face both fundamental and technical issues due to the difference in representational logic between computer graphics and knowledge-based systems. We present a framework for such an integration illustrated by a first prototype.

[top]


Papers: Gesture-and-Sketch-Based Interfaces
Wednesday, January 31st, 4:30 pm - 6:00 pm

 

Tracking of Deformable Human Hand in Real Time as Continuous Input for Gesture-based Interaction (ACM Digital Library Link)
Xiying Wang (Intelligence Engineering Lab& Laboratory of Computer Science, Institute of Software, Chinese Academy of Sciences)
Xiwen Zhang and Guozhong Dai (Intelligence Engineering Lab, Institute of Software, Chinese Academy of Sciences)

Abstract
Gesture input is a natural and effective interactive model. The tracking of deformable hand gesture is a very important task in gesture-based interaction. A novel real-time tracking approach is proposed to capture hand motion with single camera. It combines the characters of model-based and appearance-based method. The presented approach achieves auto-initialization by posture recognition and matching with image features. It solves the problem of interference among fingers successfully by the integration of K-Means clustering and Particle Filters. Moreover, tracking detection realizes resumption from tracking failure and automatic update of hand model. Experiments show that, the proposed method can achieve continuous real-time tracking of deformable hand gesture, and meets the requirements for gesture-based Human-Computer Interaction.

Crossmodal Error Correction of Continuous Handwriting Recognition by Speech (ACM Digital Library Link)
Xiang Ao, Guozhong Dai, Hongan Wang (Intelligence Engineering Lab & Laboratory of Computer Science, Institute of Software, Chinese Academy of Sciences)

Abstract
In recognition-based user interface, users' satisfaction is determined not only by recognition accuracy but also by effort to correct recognition errors. In this paper, we introduce a crossmodal error correction technique, which allows users to correct errors of Chinese handwriting recognition by speech. The focus of the paper is multimodal fusion algorithm supporting the crossmodal error correction. By fusing handwriting and speech recognition, the algorithm can correct errors in both character extraction and recognition of handwriting. The experimental result indicates that the algorithm is effective and efficient. Moreover, evaluation also shows the correction technique can help users to correct errors in handwriting recognition more efficiently than the other two error correction techniques.

NaturalDraw: Interactive Perception Based Drawing for Everyone (ACM Digital Library Link)
Yasser Mohammad and Toyoaki Nishida (Kyoto University)

Abstract
Drawing is a very natural activity of humans, and, despite the wide variety of drawing systems available on computers to-day, most of those systems lack the naturalness of the pencil-paper system. In this paper we present a new drawing inter-face that is easy for the human to use in a more natural way than the existing drawing interfaces. The proposed system is based on the Interactive Perception paradigm we developed for interfacing social robots to the humans.

[top]