The Past's Digital Presence: Some Reflections, Part II

After last week’s class discussion, I thought it appropriate to post more on the Yale conference, particularly session IV, Theorizing the Digital Archive, as most of the presentations were at least peripherally-related to issues that got brought up in our class discussion on semantic KOS for cultural heritage. 

Regarding the idea brought up in class about the artists that resist categorization (be it from situating themselves as anti-establishment or as thinking of their work as something other than “art”) and the choice to include them or not in a  cultural or art ontology: I think I may have vaguely mentioned one of the PDP presentations as relevant.  In fact, it was Stewart Campbell‘s “Eugène Atget & The Digital Archive.”  His basic premise: MoMA canonized Atget’s photographs as art photography by selecting only limited number of the works in its Atget collection for its digital collection and thus creating a selction bias–flawed works are not represented, e.g., animal photographs criticized by Berenice Abbott.  Furthermore, Atget produced his work for commercial purposes, and did not position himself as an artist.  Further information can be found in Campbell’s abstract (scroll down).

Related to our discussion of ontologies last week was Alexandre Monnin’s session IV presentation, “What is a Tag: Digital Artifacts as Hermeneutical Devices.”  Those curious about the relationship of philosophy to the Internet, specifically ontologies, may want to check out Monnin’s thesis proposal.  At one point referencing George Lakoff in the idea that classification is intimately related to power, Monnin examined the idea and function of the tag and an its application as an interpretive device.  He began by defining a tag as a “digital equivalent of a real-life tag–a blank space with digital string attaching it to an item.”  This definition in place, he continued to argue that many tags on the Web aren’t tags, but keywords or authorized vocabularies–an interesting point to us library types who typically do strive to use such applications in a more standardized and consistent way.   Richard Newman’s MOAT (Meaning of the Tag) ontology was also discussed, as was Dbpedia, and commontag.org.  More information and links can be found on Monnin’s webpage.

Hi class. Here are my notes on ontologies. Please feel free to expand:

De Ridder, J. (2007). The immediate prospects for the application of ontologies in digital libraries. Knowledge Organization, 34(4), 227-46.

Ontologies are systems created by humans to make human knowledge readable by machines by defining human vocabulary and concepts and the relationships between them. Instead of merely allowing keyword searching or simple information retrieval, ontologies allow computers to answer a query much more like a human would, by reformulating queries, understanding relationships between concepts and considering what the user is really looking for and how they want to see information, leading to more precise search results. That’s A.I., folks!

The benefits of ontologies are:

• allows search across diverse sources which may not have the same terminology or compatible metadata.
• Allows natural language searching
• Allows tailoring of results to the user’s task and needs
• Makes more complicated resources available to laypeople

Ontologies are easiest to make when they cover only one domain (subject area) and if they don’t need to crossmap with other outside ontologies. Cross mapping is difficult and time-consuming, yet is one of the only ways to make an ontology as complete as possible. It can be automated to some level, especially if the domain’s terminology and concepts are already structured and controlled (like in science), but requires much human input to make sure that the mappings make sense and are correct. In the case of cultural heritage, each domain (and even organization) contains many biases, languages, assumptions, etc, making the job of crossmapping even more difficult. Crossmapping betwen domains requires multiple domain experts and people who know more than one domain well. Crossmapping problem are similar to metadata crosswalks problems discussed in the Woodley article, except that instead of un-uniform fields, un-uniform concepts are a problem.

Creating ontologies is time consuming and expensive! They require a huge initial investment of human time and labor, and then they must be updated and maintained to keep up with changes in thinking about, description of and relations to concepts. Before you consider developing or adapting an ontology, you have to make sure it is worth it for your organization and your users.

De Ridder suggests that ontologies will only be feasible in the business sector, but that with enough funding and interest, educational environments could crank out a few beyond protoype-level.

***The thought lab in Europeana displays ontological searching much like the example in section 1.2 of the reading. When you search of a word (I used ‘bath’), the engine offers up several meanings of the word for one to choose from—making the search result more precise, though it is not set up for queries!