With the increasing amount of information available on the Internet one of the most challenging tasks is to provide search interfaces that are easy to use without having to learn a specific syntax and let people find what they really want. We have developed a natural language query interface for one of the largest European Web-based tourism information systems Tiscover. The interface allows for posing queries either in German or English to search for accommodations throughout Austria. We have conducted a field trial by making the search interface publicly available and promoting it on the Tiscover homepage. We performed a detailed analysis of the queries that were posed by examining some statistical properties of the data. We also carried out a supervised usability study in which our natural language interface was compared with the original Tiscover accommodation search interface. As enhancements of the original natural language interface we have developed: a supporting tool for ontology engineers based on neural networks to display a map showing the vocabulary extracted from free-form text documents, which is automatically organized according to the semantic similarities between the words; an approach derived from a common term weighting technique in information retrieval to detect terms that are characteristic for particular geographical regions; a knowledge base by using associative networks to define relations between domain-intrinsic information; and an algorithm for deriving concept relations from past user interactions.
Prof. Winiwarter holds a tenured position at the Faculty of Business, Economics, and Computer Science, University of Vienna. He received an MS degree in 1990, an MA degree in 1992, and a PhD degree in 1995, all from University of Vienna. Former positions: Scientific CEO of the Electronic Commerce Competence Center 2001-2002, Scientific Director of the Software Competence Center Hagenberg 1999-2001, Guest Professor at the Research Institute for Symbolic Computation, Johannes Kepler University of Linz 1999-2001, and Visiting Researcher at the Department of Information Science, Kyoto University 1995-1997. The main research interest of Prof. Winiwarter is Human Language Technology. In addition, he also works on Electronic Business, Semantic Web, Machine Learning, Information Retrieval and Filtering, Education Systems, and Security. Prof. Winiwarter organizes the 4th International Workshop on Natural Language and Information Systems (NLIS 2004). He is a member of the steering committees of the 6th International Conference on Information Integration and Web-based Applications & Services (iiWAS 2004) and the 2nd International Conference on Advances in Mobile Multimedia (MoMM 2004). Prof. Winiwarter is a PC member for the following upcoming events: EsTAL, DaWaK 2004, WebS 2004, iiWAS 2004, and NLDB 2004.
Go! is a multi-paradigm programming language that is oriented to the needs of programming secure, production quality, agent based applications. It is multi-threaded, strongly typed and higher order (in the functional programming sense). It has relation, function and action procedure definitions. Threads execute action procedures, calling functions and querying relations as need be. Threads in different agents communicate and coordinate using asynchronous messages. Threads within the same agent can also use shared dynamic relations acting as Linda style memory stores.
In this talk we introduce the essential features of Go! illustrating the language by showing how it can be used to implement a simple multi-agent application comprising hybrid reactive/deliberative agents interacting in a simulated ballroom.
In the ballroom simulation dancer agents arrive in random and phased order. On arrival they register their desires with a directory server and subscribe for details of the desires of any other dancer of opposite sex. A band periodically announces the start of a dance and the end of a dance. During each dance interval, the dancers negotiate to enter into joint commitments to dance a particular type of dance (e.g. polka) when it is next announced, or perhaps to go to the bar. The negotiation takes into account the other dancers registered desires, the dancer's own unsatisfied desires, their current beliefs about what dances they have danced and with whom, and their current commitments. When a dance is announced for which they have a commitment, they 'dance' or 'go to the bar'.
Each dancer is implemented as three concurrently executing threads - one to process messages from the directory server when a new dancer of opposite sex arrives; one to negotiate, one to listen to the band and dance. The threads within an agent communicate and coordinate using belief, desire and intention Linda style stores. The threads in different agents communicate using messages. We believe such a multi-threaded agent architecture represents a powerful and natural style of agent implementation, for which Go! is well suited.
Paper linked with the talk at http://www.doc.ic.ac.uk/~klc/dalt03.html