University of Technology, Sydney

Staff directory | Webmail | Maps | Newsroom | What's on

QCIS Seminar: Overcoming the brittleness of classical logic for knowledge representation - Dr Steven Schockaert, Cardiff Univ, UK

Presenter: Dr Steven Schockaert, Cardiff University, UK (invited by A/Prof Sanjiang Li)

Abstract: Recent advances in the field of information extraction have made it possible to acquire large logical domain theories from information on the web. As a result, a need has grown for approaches that can robustly handle the uncertainty, vagueness, subjectivity and context-dependence of such theories. Within the area of knowledge representation, a large number of formalisms have already been proposed to deal with some of these issues. Broadly these approaches can be categorised as qualitative or quantitative, and as uncertainty or similarity focused. For example, Bayesian networks and Markov logic can be seen as quantitative approaches for handling uncertainty, while System P and related approaches to default reasoning are based on a qualitative model of uncertainty. Fuzzy logic and case based reasoning can be seen as quantitative approaches to similarity based reasoning, but qualitative approaches to similarity based reasoning have not yet received much attention.


Crucially, none of the existing approaches offers a sufficiently general and robust solution to the problem of reasoning about noisy domain theories from the web. In this talk I will argue that the following three challenges need to be addressed to overcome the brittleness of classical logic as a core language for knowledge representation:


1.         A qualitative theory of similarity based reasoning needs to be developed in which plausible conclusions can be drawn in cases where knowledge bases are inconsistent or incomplete. Indeed, it is well known that humans often resort to information about similar or analogous cases when dealing with unfamiliar situations, and they do so without referring to numbers.


2.         Similarity and uncertainty based approaches need to be integrated, as they target complementary aspects of commonsense reasoning, eg. knowledge bases may at the same time be imprecise (which can only be modelled given some notion of similarity) and uncertain.


3.         Quantitative and qualitative methods need to be combined. Accurate quantitative models can usually only be derived if high-quality training data is available in abundance. Especially for complex models, eg. those combining uncertainty and similarity, this condition is typically not satisfied. Qualitative models are applicable in a wider context and are generally easier to interpret by humans, but they often yield conclusions which may be too cautious.


I will highlight recent developments on these challenges, and emphasise their relevance in the context of information extraction from the web.


Brief bio: Steven Schockaert studied computer science at Ghent University, where he also defended his PhD thesis in 2008 on the topic, "Reasoning about fuzzy temporal and spatial information from the web". This thesis was awarded with the 2008 ECCAI Artificial Intelligence Dissertation Award and the IBM Belgium Prize for Computer Science. After his PhD, he obtained a postdoctoral fellowship from the Research Foundation - Flanders. In September 2011 he became a lecturer at Cardiff University, where he currently works. His current research interests are largely centred around three areas: multi-valued logics, commonsense reasoning, and geographic information retrieval. In the area of multi-valued logics, he is mainly working on ukasiewicz logic (eg. automated reasoning and complexity) and on extensions of answer set programming, a form of logic programming based on the notion of stable models. In the area of commonsense reasoning, he looks at qualitative techniques for managing imperfect knowledge bases, including possibilistic logic and qualitative approaches to similarity-based reasoning. Finally, in the area of geographic information retrieval, his main focus is on acquiring geographic information from semi-structured web resources.


12 September 2013
11:00 - 12:15
City - Broadway CB10 Level 3, Room 330
All Welcome
Barbara Munday

Back to list of Past Events