Peter W. Eklund

Embedding knowledge in Web documents

Abstract The paper argues for the use of general and intuitive knowledge representation languages (and simpler notational variants, e.g. subsets of natural languages) for indexing the content of Web documents and representing knowledge within them. We believe that these languages have advantages over metadata languages based on the Extensible Mark-up Language (XML). Indeed, the retrieval of precise information is better supported by languages designed to represent semantic content and support logical inference, and the readability of such a language eases its exploitation, presentation and direct insertion within a document (thus also avoiding information duplication). We advocate the use of Conceptual Graphs and simpler notational variants that enhance knowledge readability. To further ease the representation process, we propose techniques allowing users to leave some knowledge terms undeclared. We also show how lexical, structural and knowledge-based techniques may be combined to retrieve or generate knowledge or Web documents. To support and guide the knowledge modeling approach, we present a top-level ontology of 400 concept and relation types. We have implemented these features in a Web-accessible tool named WebKB 2 , and show examples to illustrate them.

Knowledge retrieval and the World Wide Web

The Web is currently a distributed mass of simple hypertext documents. The authors present a new tool, WebKB that interprets semantic statements stored in Web-accessible documents. WebKB lets its users combine lexical, structural and knowledge-based techniques to exploit or generate Web documents. WebKB is an ontology server and directed Web robot.

Browsing Semi-structured Web Texts Using Formal Concept Analysis

Query-directed browsing of unstructured Web-texts using Formal Concept Analysis (FCA) confronts two problems. Firstly on-line Web-data is sometimes unstructured and any FCA-system must include additional mechanisms to structure input sources. Secondly many online collections are large and dynamic so a Web-robot must be used to automatically extract data. These issues are addressed in this paper. We report on the construction of a Web-based FCA system for browsing classified advertisements for real-estate properties. Real-estate advertisements were chosen because they are typical of semi-structured textual information sources accessible on the Web. Furthermore, the analysis of real-estate data using FCA is a classic example used in introductory courses on FCA. However, unlike the classic FCA real-estate example, whose input is a structure relational database, we automatically mine Web-based texts for their structure.

Scalability in Formal Concept Analysis

Formal Concept Analysis is a symbolic learning technique derived from mathematical algebra and order theory. The technique has been applied to a broad range of knowledge representation and exploration tasks in a number of domains. Most recorded applications of Formal Concept Analysis deal with a small number of objects and attributes, in which case the complexity of the algorithms used for indexing and retrieving data is not a significant issue. However, when Formal Concept Analysis is applied to exploration of a large numbers of objects and attributes, the size of the data makes issues of complexity and scalability crucial.

A Contextual-Logic Extension of TOSCANA

The aim of this paper is to indicate how TOSCANA may be extended to allow graphical representations not only of concept lattices but also of concept graphs in the sense of Contextual Logic. The contextual-logic extension of TOSCANA requires the logical scaling of conceptual and relational scales for which we propose the Peircean Algebraic Logic as reconstructed by R. W. Burch. As graphical representations we recommend, besides labelled line diagrams of concept lattices and Sowa’s diagrams of conceptual graphs, particular information maps for utilizing background knowledge as much as possible. Our considerations are illustrated by a small information system about the domestic flights in Austria.

Analyzing an Email Collection Using Formal Concept Analysis

We demonstrate the use of a data analysis technique called formal concept analysis (FCA) to explore information stored in a set of email documents. The user extends a pre-defined taxonomy of classifiers, designed to extract information from email documents with her own specialized classifiers. The classifiers extract information both from (i) the email headers providing structured information such as the date received, from:, to: and cc: lists, (ii) the email body containing free English text, and (iii) conjunctions of the two sources.

CEM - A Program for Visualization and Discovery in Email

This paper presents a lattice metaphor for knowledge discovery in electronic mail. It allows a user to navigate email using a lattice rather than a tree structure. By using such a conceptual multi-hierarchy, the content and shape of the lattice can be varied to accommodate queries against the email collection. This paper presents the underlying mathematical structures, and a number of examples of the lattice and multi-hierarchy working with a prototypical email collection.

Algorithms for Creating Relational Power Context Families from Conceptual Graphs

This paper reports on the implementation of Willes algorithm for creating power context families from conceptual graphs. It describes how to extend FCA tools, such as TOSCANA, to accommodate the expressibility of CGs. It further reports on a tool where Willes Algorithm, along with an algorithm to create SQL queries from power context families, has been implemented and applied. These algorithms are given a detailed presentation in this paper and their complexity provided.

A measurement theory perspective for MCDM

In this paper, we investigate the meaningfulness of conclusions drawn using the Sugeno (1974) and Choquet integrals as aggregation operators in multicriteria decision making. After introducing measurement theory and fuzzy integration in the MCDM setting, we point out that there is a limit to the comparison of alternatives that one can make, and we indicate what are the meaningful conclusions available with fuzzy integrals.

Embedding Knowledge in Web Documents: CGs versus XML-based Metadata Languages

The paper argues for the use of general and intuitive knowledge representation lajiguages for indexing the content of Web documents and representing knowledge within them. We believe these languages have advantages over metadata languages based on the Extensible Mark-up Language (XML). Indeed, the representation and retrieval of precise information is better supported by languages designed to represent semantic content and support logical inference, and the readability of such a language eases its exploitation, presentation and direct insertion within a document.To further ease the representation process, we propose techniques allowing users to leave some knowledge terms undeclared. We illustrate these ideas with WebKB, a precision-oriented information retrieval/annotation tool, and show how lexical, structural ajid knowledge-based techniques may be combined to retrieve or generate knowledge or Web documents. Finally, to overcome the scalability problems of storing knowledge within Web documents, we propose some ideas for scalable and cooperatively built knowledge repositories.