General Introduction to Topic Mapping

Michel Biezunski, High Text

Email:michel@hightext.com Web: www.hightext.com

SGML Europe 97, May 97, Barcelona

Contents

• Executive summary
• Background
• User requirements
• Conceptual description
• Technical description

Executive Summary

Management tool

• relevant interconnections used to navigate.
• various semantic levels applicable
• unobstrusiveness

Electronic equivalents for traditional navigational devices

• indexes
• glossaries
• thesauri
• catalogs
• cross-references

Designed to facilitate

• document evolution
• maintenance
• interchange.

Representation of database schemas

• Entity-relationship models
• Object-oriented modeling
• Multidimensional views on information

Objective

Provide a standardized way to interchange navigational information

Background

• Initiated within the Davenport Group (1991-1993) as SOFABED (acronym for Standard Open Formal Architecture for Browsable Electronic Documents)
• Further developed as part of the CApH (Conventions for the Applications of HyTime), an activity chaired by Steven R. Newcomb and sponsored by the GCA-RI (GCA Research Institute).
• Since October 1996, adopted by ISO WG8 as a project for a new international standard (ISO/IEC 13250).
• Two co-editors are Martin Bryan (SGML Centre, UK) and Michel Biezunski (High Text, France).

User requirements

• Control of navigational information
• Intermediary step between automatic indexing and document structuring
• Interchange of navigational information
• Management of cross-references

An SGML-based architecture

• The Topic Map specification is a meta-DTD
• Set of HyTime-inherited architectural forms, i.e. templates used for building SGML DTDs.

Current state: 3 architectural forms:

• Semantic assignment: defines a topic by linking every portion of information having to do with a particular topic.
• Semantic title, which assigns a name to a given topic.
• Topic relation, which relates multiple topics together through a relationship described as a link.

Topic Navigation Maps: An ISO standard (ISO 13250) due end of 1998

Planned extensions to the current model

• multilingual navigation
• validity of information in time
• mapping to entity-relationship schemes and generic queries for databases
• connection of multiple topic maps
• + ???

The market for Topic Maps

• When information has high value.
• No change required to legacy data
• User controlled information retrieval
• Adapted to the Web
• Individual knowledge bases

The competitive advantage

• Easy maintenance of links.
• End of cross-reference nightmare.
• Humans can focus on semantics of linking, while machines will do the addressing job.

Technical Description

The Topic Map Architectural Forms

Architectural forms are formally defined in the AFDR (Architectural Form Definition Requirements) annex of 10744 (2nd edition in preparation).

Topic Maps are based on a separation between linking and locating.

• Linking : semantics of the relationships.
• Locating : addresses of the information objects that may be used as anchors in links.

The most general template to describe Topic Maps is the HyTime representation (hyperlinks module). A significant subset of the architecture can be described using XML extended links.

Independent links

Link constructs are independent of each of their anchors.

• Non-independent link :"I am linked to X, click on me to go there"
• Independent link: "A, B, C and D happen to be linked together".

Main advantage: links can be stored in databases.

The Semantic Assignment (or Topic)

Topics are specialized link elements. They express a certain kind of relationship, and link element instances can be grouped by type. Anchors are grouped relatively to the role they play in the relationship ("anchor roles"). Other attributes are added: "mnemonic" (attribute used for example for alphabetic sort), "semantic universe" (used to differentiate domains of knowledge). Other filtering attributes will be added in a future version of the architecture, such as language, validity, etc.

Topics can have zero or several titles. A topic without a title can be interesting for describing pure navigation (such as in cross-references), a topic can also have several titles (i.e. equivalent forms, such as "art museum" and "museum of art").

Topics can be used for the representation of indexes, glossaries, and cross-references. They play the same role as index term. A glossary is the result of a query containing topic titles, followed by the content of the anchors whose anchor role is "definition". They can be used to represent cross-references, because a cross-reference usually means: to know more about the current topic, see also the next anchor related to this topic.

The topic relationship

Topic relationships are links to links: they are used to relate topics together. The nature of the role played by the relation is left to the application architect. It is possible to describe all kinds of relations existing in a thesaurus without limits on the semantics of each of the relationship.