Local Government New Zealand is considering the use of topic maps as a tool to manage public access to district plans and other documents.
LGNZ, whose brief is to help develop best practice in local government administration, first looked at the topic maps area about two years ago, staff at Victoria University’s New Zealand Electronic Text Centre (NZETC) say, but found there was not enough local expertise to embark on serious implementation.
Now there is a fund of such knowledge and experience, particularly in NZETC, LGNZ is likely to have another go at evaluating and possibly using the idea.
The topic map is a way of indexing the data in a collection of often dissimilar documents allowing a searcher to identify differently-worded references to the same topic and to move along a "link" from one topic to a related topic if their original search was not quite on-target, or they wish to expand their line of enquiry.
A consultant who works with the consortium says topic maps are definitely "a subject of interest" there again. But LGNZ communications officer Kirsty Anderson plays down interest in specific techniques. "At the moment, we are looking at options to review our records management," she says. We have looked at what several suppliers have to offer, but have not committed to any specific product."
Mike Manson, of Palmerston North City Council and ALGIM (the Association of Local Government Information Managers), says a "thesaurus" project is being discussed extensively among councils. This will gather all references in documents to an object, place, function, task and so forth, however worded, and link them to a common standard term. This is part of what topic maps can do, but Manson says he is not familiar with the phrase.
A topic map is more discriminating than a typical search engine, in that it references useful occurrences of the topic, rather than simply occurrences of a phrase. NZETC’s Elizabeth Styron characterises the difference: if you searched the IRD website for the name of a form, you’d probably get every reference to the form in any document, but you may well not find the form itself. The topic map will have a direct pointer to the form.
Because they deal in concepts and not words, topic maps can sit over a heterogeneous set of documents differently arranged, including differing degrees of detail and stored in different data formats on dissimilar computer systems, she says.
A district plan and related documents for Auckland would, for example, be a lot larger and likely to have much finer detail than similar documents for Gore; but with a topic map referencing both, a citizen of either town could, through the same web interface, find, say, a checklist of permissions needed to build a new extension to a house.
Generating a topic map is first a matter of building an ontology — a collection of the important entities in the document collection. Some of these may be data entities, such as a web page, PDF document, book, page or image, and some may be real-world entities, such as a person, task, date or place.
Relations can be among or across these types; a page is within a book; a person performs a task; an image is contained in a document and depicts a person. Thus a searcher can “surf” through connected topics.
Different views may be put on the same collection of documents; a local authority may reference subjects through a topic map in a different way for a planner and an accountant, or a museum may provide access to more complex descriptions of items it holds for an adult researcher than for a child.
Words likely to indicate an entity of a specific type identified in the map (for example all personal names) can be tagged using XML. A “harvester” program primed with the ontology can then go through the documents and build the map automatically, though some of the metadata may need tuning manually, Styron says.