The concept of Semantic Web (Berners, 2001) introduces a new form of knowledge representation – an ontology. An ontology is a partially ordered set of words and concepts of a specific domain, and allows for defining different kinds of relationships existing among concepts. Such approach promises formation of an environment where information is easily accessible and understandable for any system, application and/or human. Hierarchy of concepts (Yager, 2000) is a different and very interesting form of knowledge representation. A graph-like structure of the hierarchy provides a user with a suitable tool for identifying variety of different associations among concepts. These associations express user’s perceptions of relations among concepts, and lead to representing definitions of concepts in a human-like way. The Internet becomes an overwhelming repository of documents. This enormous storage of information will be effectively used when users will be equipped with systems capable of finding related documents quickly and correctly. The proposed work addresses that issue. It offers an approach that combines a hierarchy of concepts and ontology for the task of identifying web documents in the environment of the Semantic Web. A user provides a simple query in the form a hierarchy that only partially “describes” documents (s)he wants to retrieve from the web. The hierarchy is treated as a “seed” representing user’s initial knowledge about concepts covered by required documents. Ontologies are treated as supplementary knowledge bases. They are used to instantiate the hierarchy with concrete information, as well as to enhance it with new concepts initially unknown to the user. The proposed approach is used to design a prototype system for document identification in the web environment. The description of the system and the results of preliminary experiments are presented.
TopIntroduction
Amount of information available on the Internet creates a number of challenges for the text identification processes. Users should be able to find web pages containing text/documents that belong to a category of interest without missing too many of them. Development of effortless and efficient ways of finding such documents is of critical importance.
A number of methods addressing the issue of identifying relevant documents have been developed recently. Majority of them are built based on different Machine Learning techniques. Those methods “see” documents as vectors of weighted terms (words). Many systems constructed for text categorization purposes induce vectors that are characteristic for each category. In a nutshell, a categorization process takes place via comparison of those characteristic vectors with vectors representing documents.
For a person, identification of documents that belong to a specific category, as well as documents that are related to it, is an everyday activity. A person finds a document based on concepts related to their category of interest. The set of concepts is like a network. Individual concepts are linked among themselves, and the links represent different relationships that can exist between those concepts. Some of possible relationships are: meronymy – when a concept C’ is part of C”, holonymy – a concept C’ has C” as a part, hyponymy (troponymy) – in the case when C’ is subordinate of C”, or synonymy – if C’ denotes the same as C”. A document identification process is performed via exploration of concepts and links between them. If one concept is found in a document, it “activates” concepts linked to it, and the document is checked if it contains those activated concepts. Eventually, more and more evidence is collected towards the statement that the document belongs to a category of interest. When some threshold value is reached the identification process is stopped. The idea of mimicking such an identification process is explored in the paper.
Introduced in May 2001, the concept of Semantic Web (Berners, 2001), seen as an extension of the current web, defines an environment in which information is given a well-defined meaning. The Semantic Web is a place where machines can analyze all the data on the Web (Fensel, 2003) (Antoniou, 2004). A common element of all of those definitions is a reference to a new method of representing data. A new representation of resources on the web is based on usage of ontology. An ontology is a formal, explicit specification of a shared conceptualization (Gruber, 1993). It is a set of well-defined classes that describe data models in a specific domain. (The term class will be used in this work to represent a concept defined in ontologies. Therefore, the term concept as used in hierarchies of concepts, and the term class as used in ontologies mean the same thing – a concept. The term category represents a set of entities – web pages in this work – that are grouped together due to their relevance to the same concept.) Together with their individuals (instances of classes), ontologies work as knowledge vehicles to express individual facts (Scott, 2002). This new representation of knowledge introduced to the web environment brings new possibilities of utilization of information (Marin, 2004; Sanchez, 2006).