Evolutions in Subject Searching

Slides are available here for Evolutions in Subject Searching: the Use of Topic Maps in Libraries with Steve Newcomb, co-founder of topicmaps.org and a co-author of a topic maps standard, and Patrick Durusau, on the board of TEI as well as involved in other markup standards organizations (didn’t catch them all).

I had assumed this session would be about things like AquaBrowser, but in fact, it was about an approach to representing knowledge in an expandable, shareable data structure.

Apparently topic maps and XML-based topic maps are big in a whole other context outside the library world, wherever people are managing industrial quantities of knowledge or documents. Basically, this talk addressed the structural underpinnings of one way to do a semantic map of a domain, and be able to construct a crosswalk to another domain using the same map. Sounds very labor-intensive.

One thing I learned is that in the world of topic maps, everything is a subject. This is not Anglo-American Cataloging Rules, here. Any data element of any kind can be a “subject” in a topic map.

Did you know there is a LITA Topic Maps IG? If you’re at LITA Forum, catch Suellen Stringer-Hye to ask about it. They will also be meeting at Midwinter.

Steve Newcomb

This speaker went through a number of slides that seemed to jump around the topic, as I adjusted to the fact that I’d walked in the room knowing nothing about this kind of topic map:

  • The most basic thing about topic maps: one subject for each location
  • The subject doesn’t need to be identified in any specific way, but must be identified
  • Using topic mapping, you can create bridges or “wormholes” between heterogeneous information sources and representations, or universes of discourse
  • A subject is defined by key-value pairs
  • XTM: XML Topic Mapping
  • Related organizations, companies, and conferences:
    • www.ieml.org
    • www.versavant.org
    • Atlas Elektronik
    • www.cit.de topic maps for municipal information for city of Stuttgard
    • US Department of Energy uses topic maps for asset management, weapons secrets classification
    • Dutch Tax and Customs uses topic maps to identify duplicate sources of publications and select the free or lower-cost version, rather than paying vendors twice for the same information
    • Extreme Markup Languages annual conference – primary industrial conference
    • TMRA Topic Maps Research and Applications – primary academic conference

Patrick Durusau

Different people define, or identify, subjects in different ways. How to capture different identifications? How to reuse mapping of identifications?

Where this problem may become critical is, for example, when medical terminology changes. Existing studies may become essentially unfindable, or hard to find without chasing down a trail of see-also references. A teaching hospital’s own study on a medical condition was too old to be found with current medical terms; the patient dies.

Subject identification by term or by unique identifier works within slow-changing areas and within a single area where everyone uses the same vocabulary or identifier set — otherwise, it breaks down.

Topic maps use a representative (proxy) for a subject that can contain multiple identifications for the same subject. (Example: Mark Twain and all his many noms de plume.)

All keys in a topic map are references to proxies. The topic map contains a legend. The topic map is self-documenting and therefore can be shared.

Q & A:
Q: Libraries already have such detailed subject indexing and cross-referencing. Why would it be worth it for a library to go to the trouble to create a topic map?

A: Because other libraries or institutions can contribute or share, especially when you are trying to provide a single point of access into multiple semantic tagging systems (such as LCSH and a museum database).

Q: Why didn’t you talk about the visual displays of topic maps?

A: This talk is not directly about the user interfaces; the topic mapping initiatives are concerned with both the interface and the underlying data representations.

Q: Does the topic map tell only about synonyms? What does it tell us about related terms, broader terms, narrower terms?

A: In topic maps, relationships themselves are also subjects, with properties identifying what kind of relationship.