Marko Balabanovic

I'm currently at lastminute.com (now the European arm of Travelocity) in London developing personalisation and recommendation systems for the consumer travel and entertainment market.

In the last few years I've helped set up a couple of startups in London: Cellectivity, creating m-commerce mobile phone applications from existing web e-commerce sites using web wrapper technology, and flutter.com - now Betfair, the world's first and still biggest person-to-person betting exchange.

Contact: mb at balabanovic dot mailshell dot c o m

Research Projects

Top - Recommender Systems - Communication Appliances - Mobile Robots

Recommender Systems

Machine Learning, Artificial Intelligence, Information Retrieval, Human-Computer Interaction, Personalisation

The goal of a recommender system is to learn about a population of users in order to provide increasingly accurate recommendations - of books, movies, news articles, web pages, or any kind of item. The work described here started at Stanford University during the early days of the Web in 1994, with intelligent agents that discovered new web pages of interest to users. In 1997 the multiagent Fab system was released for public use, combining both content-based and collaborative filtering, and in 1998 the Slider user interface was introduced to infer feedback from users' drag/drop/click actions. The research was done in the context of several research groups: the Stanford Digital Library project, the Multiagent group and the HCI group.

Learning to Surf contains the most complete explanations of the Slider, Fab and Lira systems described below. The research was done with help from Yoav Shoham (PhD advisor), and Terry Winograd & Rob Barrett (reading committee members).

Learning to Surf: Multiagent Systems for Adaptive Web Page Recommendation
PhD Thesis, Stanford University Department of Computer Science, completed March 1998
- Available as Stanford University Technical Report CS-TR-98-1605 (1MB PDF)

Slider was an interface to a system that recommended news articles. One of the contributions of this research was an interaction design that inferred users' document preferences by monitoring their drag-and-drop actions. Such implicit feedback is less burdensome to users than having to explicitly rate or rank news articles. A second contribution of the research was a mechanism for users to easily define multiple topics of interest and control the proportions between them.

An Interface for Learning Multi-topic User Profiles from Implicit Feedback
AAAI-98 Workshop on Recommender Systems, July 1998, Madison, Wisconsin
- Earlier version: Stanford Digital Library Project Working Paper SIDL-WP-1998-0089
The "Slider" Interface
IBM interVisions #11, February 1998

Fab was an adaptive, multiagent system for recommending Web pages. This 2-year project dealt with issues such as combining content-based and collaborative recommendation, scaling up the number of users for a fixed number of searching agents, multiagent architectures, the exploration versus exploitation tradeoff when making recommendations, and machine learning techniques for Web search and recommendation.

Exploring versus Exploiting when Learning User Models for Text Recommendation
User Modeling and User-Adapted Interaction 8(1), 1998 (special issue on Machine Learning for User Modeling)
- Earlier version: Stanford Digital Library Project Working Paper SIDL-WP-1997-0067
Fab: Content-based, Collaborative Recommendation
with Yoav Shoham
Communications of the ACM 40(3), March 97
- Full paper (484K pdf)
  (Requires access to ACM's digital library)
An Adaptive Web Page Recommendation Service
First International Conference on Autonomous Agents, February 1997, Marina del Rey CA
- Earlier version: Stanford Digital Library Project Working Paper SIDL-WP-1996-0041
Some ideas from Fab were implemented commercially as part of the Imana CommonQuest knowledge management product.

LIRA was a precursor to Fab, a simple single-user single-agent recommendation system.

Learning Information Retrieval Agents: Experiments with Automated Web Browsing
(Extended Abstract)
with Yoav Shoham
AAAI-95 Spring Symposium on Information Gathering from Heterogenous, Distributed Environments
- Published in AAAI Tech report SS-95-08
An Adaptive Agent for Automated Web Browsing
with Yoav Shoham and Yeogirl Yun
September 1995
- Stanford University Technical Report CS-TN-97-52 (PDF)
- Stanford Digital Library Project Working Paper SIDL-WP-1995-0023

Top - Recommender Systems - Communication Appliances - Mobile Robots

Communication Appliances

Human-Computer Interaction, Computer-Mediated Communication, Multimedia Authoring and Communication, Storytelling and Narrative

The Homer project at Ricoh Innovations was an initiative to create new interfaces and infrastructure for exploring, creating, finding, remembering and experiencing information.

StoryTrack was a hand-held electronic photo album that would capture audio stories users told while viewing pictures. Its business counterpart, TouchVerse, was a Windows application for easy creation of communications comprising audio and references to documents.

Storytelling with Digital Photographs
with Lonny Chu and Greg Wolff
Conference on Human Factors in Computing Systems (CHI 2000), April 2000, The Hague, Netherlands
- Full paper (PDF)
Multimedia Chronicles for Business Communication
Hawaii International Conference on Systems Sciences, January 2000, Maui HI
- Full paper (PDF)
Method and an apparatus for visual summarization of documents
with Dar-Shyang Lee
US Patent 6895552, May 2005
Game console based digital photo album
with Greg Wolff and Makoto Tanaka
US Patent 6833848, December 2004
Method and system for electronic message composition with relevant documents
with Greg Wolff
US Patent 6782393, August 2004
Method and apparatus for generating visual representations for audio documents
US Patent 6624826, September 2003
Method and apparatus for recording and playback of multidimensional walkthrough narratives
US Patent 6480191, November 2002
Note: Further patents pending

The Martian Communicator was a project done as a summer intern at what was then called Rank Xerox EuroPARC in Cambridge, working with Mik Lamming, Mike Flynn and Marge Eldrige. The idea was to unify various kinds of messaging into a simple, small, mobile device.

The Satchel was a mobile device for storage of documents. These could be transmitted to or received from nearby appliances such as printers, scanners and fax machines or exchanged with other users. In 1994 this was implemented this on the Apple Newton platform.

The Newton Satchel Movie (Quicktime, 5 min, 44MB)

Top - Recommender Systems - Communication Appliances - Mobile Robots

Mobile Robots

Artificial Intelligence, Robotics, Pattern Recognition

The SCIMMER project ("Sarah, Craig, Illah and Marko's Most Excellent Robot") resulted in the winning entry to the AAAI-93 mobile robot competition. SCIMMER could sing, dance and escape from cluttered offices at high speed.

The Real-World Navigator
with Craig Becker, Sarah K. Morse and Illah R. Nourbakhsh
AIAA/NASA Conference on Intelligent Robots in Field, Factory, Service and Space, March 1994, Houston, Texas
- Full paper (2.5MB postscript)
The Winning Robots from the 1993 Robot Competition
with Illah Nourbakhsh, Sarah Morse, Craig Becker, Erann Gat, Reid Simmons, Steven Goodridge, Harsh Potlapalli, David Hinkle, Ken Jung, and David Van Vactor
AAAI Magazine 14(4) Winter 1993, 51-62
- Abstract
- Full paper (PDF)

AIbots was a project guided by Barbara Hayes-Roth and Nils Nilsson, with the goal of unifying low-level robot control with higher-level AI architectures.

A Domain-Specific Software Architecture for Adaptive Intelligent Systems
with Barbara Hayes-Roth, Karl Pfleger, Philippe Lalanda, Phillipe Morignot
IEEE Transactions on Software Engineering 21(4), April 1995
- Earlier version: Stanford University Knowledge Systems Laboratory Report KSL-94-11
Plans and Behavior in Intelligent Agents
with Barbara Hayes-Roth, Philippe Lalanda, Philippe Morignot, Karl Pfleger
Stanford University Knowledge Systems Laboratory Report KSL-93-43
- Abstract
- Full Paper (Postscript)

Top - Recommender Systems - Communication Appliances - Mobile Robots

	Marko Balabanovic I'm currently at lastminute.com (now the European arm of Travelocity) in London developing personalisation and recommendation systems for the consumer travel and entertainment market. In the last few years I've helped set up a couple of startups in London: Cellectivity, creating m-commerce mobile phone applications from existing web e-commerce sites using web wrapper technology, and flutter.com - now Betfair, the world's first and still biggest person-to-person betting exchange. Contact: mb at balabanovic dot mailshell dot c o m Research Projects Top - Recommender Systems - Communication Appliances - Mobile Robots Recommender Systems Machine Learning, Artificial Intelligence, Information Retrieval, Human-Computer Interaction, Personalisation The goal of a recommender system is to learn about a population of users in order to provide increasingly accurate recommendations - of books, movies, news articles, web pages, or any kind of item. The work described here started at Stanford University during the early days of the Web in 1994, with intelligent agents that discovered new web pages of interest to users. In 1997 the multiagent Fab system was released for public use, combining both content-based and collaborative filtering, and in 1998 the Slider user interface was introduced to infer feedback from users' drag/drop/click actions. The research was done in the context of several research groups: the Stanford Digital Library project, the Multiagent group and the HCI group.
	Learning to Surf contains the most complete explanations of the Slider, Fab and Lira systems described below. The research was done with help from Yoav Shoham (PhD advisor), and Terry Winograd & Rob Barrett (reading committee members). Learning to Surf: Multiagent Systems for Adaptive Web Page Recommendation PhD Thesis, Stanford University Department of Computer Science, completed March 1998 Available as Stanford University Technical Report CS-TR-98-1605 (1MB PDF)
	Slider was an interface to a system that recommended news articles. One of the contributions of this research was an interaction design that inferred users' document preferences by monitoring their drag-and-drop actions. Such implicit feedback is less burdensome to users than having to explicitly rate or rank news articles. A second contribution of the research was a mechanism for users to easily define multiple topics of interest and control the proportions between them. An Interface for Learning Multi-topic User Profiles from Implicit Feedback AAAI-98 Workshop on Recommender Systems, July 1998, Madison, Wisconsin Earlier version: Stanford Digital Library Project Working Paper SIDL-WP-1998-0089 The "Slider" Interface IBM interVisions #11, February 1998
	Fab was an adaptive, multiagent system for recommending Web pages. This 2-year project dealt with issues such as combining content-based and collaborative recommendation, scaling up the number of users for a fixed number of searching agents, multiagent architectures, the exploration versus exploitation tradeoff when making recommendations, and machine learning techniques for Web search and recommendation. Exploring versus Exploiting when Learning User Models for Text Recommendation User Modeling and User-Adapted Interaction 8(1), 1998 (special issue on Machine Learning for User Modeling) Earlier version: Stanford Digital Library Project Working Paper SIDL-WP-1997-0067 Fab: Content-based, Collaborative Recommendation with Yoav Shoham Communications of the ACM 40(3), March 97 Full paper (484K pdf) (Requires access to ACM's digital library) An Adaptive Web Page Recommendation Service First International Conference on Autonomous Agents, February 1997, Marina del Rey CA Earlier version: Stanford Digital Library Project Working Paper SIDL-WP-1996-0041 Some ideas from Fab were implemented commercially as part of the Imana CommonQuest knowledge management product.
	LIRA was a precursor to Fab, a simple single-user single-agent recommendation system. Learning Information Retrieval Agents: Experiments with Automated Web Browsing (Extended Abstract) with Yoav Shoham AAAI-95 Spring Symposium on Information Gathering from Heterogenous, Distributed Environments Published in AAAI Tech report SS-95-08 An Adaptive Agent for Automated Web Browsing with Yoav Shoham and Yeogirl Yun September 1995 Stanford University Technical Report CS-TN-97-52 (PDF) Stanford Digital Library Project Working Paper SIDL-WP-1995-0023
	Top - Recommender Systems - Communication Appliances - Mobile Robots Communication Appliances Human-Computer Interaction, Computer-Mediated Communication, Multimedia Authoring and Communication, Storytelling and Narrative The Homer project at Ricoh Innovations was an initiative to create new interfaces and infrastructure for exploring, creating, finding, remembering and experiencing information.
	StoryTrack was a hand-held electronic photo album that would capture audio stories users told while viewing pictures. Its business counterpart, TouchVerse, was a Windows application for easy creation of communications comprising audio and references to documents. Storytelling with Digital Photographs with Lonny Chu and Greg Wolff Conference on Human Factors in Computing Systems (CHI 2000), April 2000, The Hague, Netherlands Full paper (PDF) Multimedia Chronicles for Business Communication Hawaii International Conference on Systems Sciences, January 2000, Maui HI Full paper (PDF) Method and an apparatus for visual summarization of documents with Dar-Shyang Lee US Patent 6895552, May 2005 Game console based digital photo album with Greg Wolff and Makoto Tanaka US Patent 6833848, December 2004 Method and system for electronic message composition with relevant documents with Greg Wolff US Patent 6782393, August 2004 Method and apparatus for generating visual representations for audio documents US Patent 6624826, September 2003 Method and apparatus for recording and playback of multidimensional walkthrough narratives US Patent 6480191, November 2002 Note: Further patents pending
	The Martian Communicator was a project done as a summer intern at what was then called Rank Xerox EuroPARC in Cambridge, working with Mik Lamming, Mike Flynn and Marge Eldrige. The idea was to unify various kinds of messaging into a simple, small, mobile device.
	The Satchel was a mobile device for storage of documents. These could be transmitted to or received from nearby appliances such as printers, scanners and fax machines or exchanged with other users. In 1994 this was implemented this on the Apple Newton platform. The Newton Satchel Movie (Quicktime, 5 min, 44MB)
	Top - Recommender Systems - Communication Appliances - Mobile Robots Mobile Robots Artificial Intelligence, Robotics, Pattern Recognition
	The SCIMMER project ("Sarah, Craig, Illah and Marko's Most Excellent Robot") resulted in the winning entry to the AAAI-93 mobile robot competition. SCIMMER could sing, dance and escape from cluttered offices at high speed. The Real-World Navigator with Craig Becker, Sarah K. Morse and Illah R. Nourbakhsh AIAA/NASA Conference on Intelligent Robots in Field, Factory, Service and Space, March 1994, Houston, Texas Full paper (2.5MB postscript) The Winning Robots from the 1993 Robot Competition with Illah Nourbakhsh, Sarah Morse, Craig Becker, Erann Gat, Reid Simmons, Steven Goodridge, Harsh Potlapalli, David Hinkle, Ken Jung, and David Van Vactor AAAI Magazine 14(4) Winter 1993, 51-62 Abstract Full paper (PDF)
	AIbots was a project guided by Barbara Hayes-Roth and Nils Nilsson, with the goal of unifying low-level robot control with higher-level AI architectures. A Domain-Specific Software Architecture for Adaptive Intelligent Systems with Barbara Hayes-Roth, Karl Pfleger, Philippe Lalanda, Phillipe Morignot IEEE Transactions on Software Engineering 21(4), April 1995 Earlier version: Stanford University Knowledge Systems Laboratory Report KSL-94-11 Plans and Behavior in Intelligent Agents with Barbara Hayes-Roth, Philippe Lalanda, Philippe Morignot, Karl Pfleger Stanford University Knowledge Systems Laboratory Report KSL-93-43 Abstract Full Paper (Postscript) Top - Recommender Systems - Communication Appliances - Mobile Robots