Intelligent Query Answering

Intelligent Query Answering

Zbigniew W. Ras (University of North Carolina, Charlotte, USA) and Agnieszka Dardzinska (Bialystok Technical University, Poland)
Copyright: © 2009 |Pages: 6
DOI: 10.4018/978-1-60566-010-3.ch166
OnDemand PDF Download:
$30.00
List Price: $37.50

Abstract

One way to make Query Answering System (QAS) intelligent is to assume a hierarchical structure of its attributes. Such systems have been investigated by (Cuppens & Demolombe, 1988), (Gal & Minker, 1988), (Gaasterland et al., 1992) and they are called cooperative. Any attribute value listed in a query, submitted to cooperative QAS, is seen as a node of the tree representing that attribute. If QAS retrieves no objects supporting query q, from a queried information system S, then any attribute value listed in q can be generalized and the same the number of objects supporting q in S can increase. In cooperative systems, these generalizations are controlled either by users (Gal & Minker, 1988), or by knowledge discovery techniques (Muslea, 2004). If QAS for S collaborates and exchanges knowledge with other systems, then it is also called intelligent. In papers (Ras & Dardzinska, 2004, 2006), a guided process of rules extraction and their goal-oriented exchange among systems is proposed. These rules define foreign attribute values for S and they are used to construct new attributes and/or impute null or hidden values of attributes in S. By enlarging the set of attributes from which queries for S can be built and by reducing the incompleteness of S, we not only enlarge the set of queries which QAS can successfully handle but also we increase the overall number of retrieved objects. So, QAS based on knowledge discovery has two classical scenarios which need to be considered: • System is standalone and incomplete. Classification rules are extracted and used to predict what values should replace null values before any query is answered. • System is distributed with autonomous sites (including site S). User needs to retrieve objects from S satisfying query q containing nonlocal attributes for S. We search for definitions of these non-local attributes at remote sites for S and use them to approximate q (Ras & Zytkow, 2000), (Ras & Dardzinska, 2004, 2006). The goal of this article is to provide foundations and basic results for knowledge-discovery based QAS.
Chapter Preview
Top

Introduction

One way to make Query Answering System (QAS) intelligent is to assume a hierarchical structure of its attributes. Such systems have been investigated by (Cuppens & Demolombe, 1988), (Gal & Minker, 1988), (Gaasterland et al., 1992) and they are called cooperative. Any attribute value listed in a query, submitted to cooperative QAS, is seen as a node of the tree representing that attribute. If QAS retrieves no objects supporting query q, from a queried information system S, then any attribute value listed in q can be generalized and the same the number of objects supporting q in S can increase. In cooperative systems, these generalizations are controlled either by users (Gal & Minker, 1988), or by knowledge discovery techniques (Muslea, 2004).

If QAS for S collaborates and exchanges knowledge with other systems, then it is also called intelligent. In papers (Ras & Dardzinska, 2004, 2006), a guided process of rules extraction and their goal-oriented exchange among systems is proposed. These rules define foreign attribute values for S and they are used to construct new attributes and/or impute null or hidden values of attributes in S. By enlarging the set of attributes from which queries for S can be built and by reducing the incompleteness of S, we not only enlarge the set of queries which QAS can successfully handle but also we increase the overall number of retrieved objects.

So, QAS based on knowledge discovery has two classical scenarios which need to be considered:

  • System is standalone and incomplete.

Classification rules are extracted and used to predict what values should replace null values before any query is answered.

  • System is distributed with autonomous sites (including site S). User needs to retrieve objects from S satisfying query q containing nonlocal attributes for S.

We search for definitions of these non-local attributes at remote sites for S and use them to approximate q (Ras & Zytkow, 2000), (Ras & Dardzinska, 2004, 2006).

The goal of this article is to provide foundations and basic results for knowledge-discovery based QAS.

Top

Background

Modern query answering systems area of research is related to enhancements of query-answering systems into intelligent systems. The emphasis is on problems in users posing queries and systems producing answers. This becomes more and more relevant as the amount of information available from local or distributed information sources increases. We need systems not only easy to use but also intelligent in handling the users’ needs. A query-answering system often replaces human with expertise in the domain of interest, thus it is important, from the user’s point of view, to compare the system and the human expert as alternative means for accessing information.

Complete Chapter List

Search this Book:
Reset