|Australasian Language Technology Workshop 2007|
Location information is available here.
Tutorial 1Presenter: Shlomo Berkovsky
Title: Web Personalization and Recommender Systems
During the last years, the quantity of potentially interesting Web-based information services has been growing rapidly and exceeds human processing capabilities. In many situations the users would like to choose among numerous alternative services, but do not have sufficient knowledge, capabilities or time to make such decisions. As such, there is a pressing need for intelligent systems that assist users while taking into account their personal needs and interests and deliver service in a way that will be most appropriate and valuable to the users. These types of systems are referred to as personalization systems.
This tutorial focuses on recommender systems, a typical example of personalization systems, which provide users with recommendations about services they may like. In these systems, generation of personalized recommendations is achieved by exploiting various knowledge sources, which store information collected during past interactions between the system and the users. Extensive research of recommender systems started over a decade ago and yielded a wide variety of recommendation techniques, such as content-based filtering, collaborative filtering, knowledge-based recommendations, utility-based recommendations and their multiple hybridizations.
This tutorial will include an overview of current recommender system applications, examples of Web-based recommender systems, a detailed overview of the most widely-used recommendation techniques, and a set of recommender systems design principles. No prior experience with recommender systems is necessary.Bio:
Shlomo Berkovsky is a Research Fellow at the Computer Science and Software Engineering Department (The University of Melbourne). His research focuses on recommender systems, user modeling and mediation of user models in recommender systems.
Tutorial 2Presenter: David Martinez
Title: Experimental method and Evaluation in Word Sense Disambiguation
Word Sense Disambiguation (WSD) has been studied as a separate NLP task since the early work on Machine Translation on the fifties. Defined as a classification task with a fixed list of senses, it has provided a common framework for experimentation on semantic phenomena. More recently, since 1998, the Senseval challenge has provided the opportunity for researchers on this area to present and evaluate their systems. These evaluations have shown that state-of-the-art systems are able to achieve around 70% accuracy for fine-grained sense distinctions, but only when hand-tagged data from the same corpus is provided. These systems are called supervised, and have well-known limitations in scalability and portability to other domains.
A different approach to WSD is the unsupervised setting, where word senses are induced directly from the corpus. Typically these systems involve clustering techniques, which group together similar examples. The evaluation of these methods is not straightforward and may involve the mapping into a known inventory or the use of goldstandard clusters.
Together with these views of the WSD problem, other related tasks have been proposed as a way to analyse the semantics of the text, such as semantic role labelling, identification of affectiveness, or lexical substitution. The goal of all these tasks is to link the text to some semantic representation. In 2007, the Senseval framework changed its name to Semeval (Semantic Evaluations) to accommodate these tasks in a common framework.
The motivation of this tutorial is to provide an overview of experimentation and evaluation for WSD and closely related tasks. We will focus on the tasks and systems presented in Senseval and the recent Semeval 2007 workshop, but will also cover other domains where WSD systems are being applied, such as biomedicine.
David Martinez is a Postdoctoral Research Fellow at the Computer Science and Software Engineering Department (University of Melbourne). His main research interests include Word Sense Disambiguation and Lexical Acquisition.
The Australasian Language Technology Workshop is being organised by ALTA, the Australasian Language Technology Association.
For any comments or questions about the workshop please contact the organizers (workshop AT alta DOT asn DOT au).