CaSTA 06 Conference Notes

Vika Zafrin, "Two Perspectives on Collaboration in the Humanities"

Vika demonstrated and talked about the Virtual Humanities Lab which has an important feature missing in many e-text projects - an annotation system that lets scholars add to the knowledge on the system. She made an interesting point about encoding. At the VHL they gave content experts the freedom to encode what they wanted and only after tried to rationalize the encoding. This makes sense in the early stages where you want content experts to focus on the semantic encoding rather than trying to develop a structure beforehand. It reminded me of how I sometimes miss COCOA and TACT - COCOA with its simple <parameter value> tagging was great for what gets called exploratory markup where you don't know what structure you are going to find. You just drop these tags in and don't worry about overlapping elements (there are not elements in COCOA markup, just parameters with different values.) The downside, as Willard pointed out in his talk (see below) is you end up with a mess that is difficult to manage. You can't change your mind easily and decide that all of these tags really ought to be some of those.

Jane Morris, "Reader's Subjective Perceptions of Lexical Cohesion and Its Implications for Computer's Interpretations of Text Meaning"

Jane Morris from Information Studies at the University of Toronto gave a paper about a study looking at how consistent readers are in their identification of clusters of words in what they read. She argued that while there is similarity between readers, there is still significant difference between the perceptions of readers. Meaning is not only in the text, but partly in the subjective reading. This isn't really a surprise, though it may be to computational linguists, what Jane did was to try to study the difference closely. What intrigued me was the idea that we should be building tools that allow interpreters to build a personal thesaurus of clusters of words under headings, which they can then use to filter and search information. Jane was arguing for reader perspective tools that embrace difference rather than try to sand it down.

Zachary Devereaux and Christopher Moore, "Shallow, Narrow, Deep and Wide: Issue Crawling for Chinese Adoption"

Devereaux and Moore, working with people like Stan Ruecker and others at Alberta have been using Issue Crawler to generate maps of networks of web sites and then trying to use them to interpret the networks. They talked about how different types of crawls (shallow vs deep) generate different results. They have an interesting hermeneutical circle problem that knowledge of the domain mapped helps them decide which maps best represent the phenomenon at the same time as the maps are supposed to help interpret the phenomenon. The representation and the represented. We had an interesting conversation about what these maps mean and how they are interpreted after their talk.

Willard McCarty, "Beyond the Word: Modelling Literary Context"

Willard gave one of his beautifully crafted talks that surveyed the types of text analysis and led to reflections on how we can draw on modelling and other scientific practices. He laid out 4 types of text analysis:

• Text encoding - analysis by careful tagging of a text
• Relational database design - the careful thinking through of how to represent a phenomenon in a relational database
• Concordancing (KWIC) - the bringing together of passages that concord in some way
• Statistical analysis of the sort John Burrows does

I'm not sure where imaging and visualization fit in here and we had a conversation about this. If by text analysis we mean ways of representation that involve analytical choices then certain types of imaging do. Visualization on the other hand could be seen as a way of showing the results of statistical analysis.

Willard went on to make a subtle argument that I can't reproduce here, but which shows a way to bridge the sciences and humanities using scientific methods (or better yet, practices) "as if" they applied to our phenomena. He took care to say that such application needs to be done carefully so as to avoid scientism. The methods only apply analogically so we will never get the sort of proof they have in the sciences, but we might in return teach the sciences about their methods. I'm sure I haven't captured the argument adequately.

I meant to ask him what the pictures in his final slides were about.

Terry Butler, "Automated Indexing Using an Existing Thesaurus: A Bridge Between Coleridge and Roget"

Terry Butler showed a neat project that is using Roget's Thesaurus (from the 1920s I believe) to help enrich the interpretative environment being created for the Coleridge Notebooks project. I'm sure there is a generalizable tool here for helping think through strange language.

Stan Ruecker, "Communicating Process with Form: Designing the Visual Morphology of the Nora Data Mining Kernels"

Stan showed some of the prototype designs being developed for the [[http://nora.lis.uiuc.edu/][NORA] project. These present collections of decisions as kernels that look like snowflakes that get more complex as they involve more decisions. It is hard to tell where this is going - the kernels are almost visual tokens that might be manipulated - each kernel representing a collection of decisions which is almost a program. I found the screens demonstrate beautiful, as anything Stan does is. It's a pity that the humanists Stan and colleagues showed the screens to were "horrified". As Stan put it, the more ludic among us like the designs, but the traditional humanists do not. What does that mean?

CaSTA 06 Conference Notes

These are conference notes on CaSTA (Canadian Symposium on Text Analysis) 06: The Breadth of Text - A Joint Computer Science & Humanities Computing Conference

See the conference site: http://www.lib.unb.ca/casta2006/

Day 1 - Thursday October 12th

Peter Shillingsburg, "Some Functions of Textual Criticism"

"My field (textual editing) presents the problem and your field (humanities computing) presents the solution"

Peter described what textual critics do partly in order to answer the question "What do I do that matters?" and partly to provide computing humanists with a sense of what critics do such that text analysis systems can support them. Shillingsburg's proposal, The Carrell(TM), provides a sense of what he imagines in a digital scholar's environment. Much of the talk however built the case for the importance of textual criticism to the university and society. The argument went something like this:

• All knowledge is recorded textually (even if also in other forms)
• Therefore the idustries of copying and disseminating text matter
• It is important to preserve the integrity of the text. This is the first function of the textual critic - proofreading
• Where texts get corrupted or there isn't a "stable" text we need critics in their second function as emenders
• Finally, and here I am not sure I heard Peter right, there is a function of teacher - bringing the rich historical, cultural, and bibliographic context to bear on the interpretation of a text. I think this is something all of us in the humanities do, not just textual scholars.

Peter also engaged questions around recent critical theory and the postmodern turn against approaches to editing that priviledged authorial text. He pointed out that that, "There are wrong ways to edit texts, but is there only one right way?" He argued for sensitivity to the way texts can be used to manipulate people and that the critic should let readers know what text is in hand and what its salient features are.

Peter started us by reminding us of the importance of the liberal arts to a university which wants to do more than collect money.

-- GeoffreyRockwell - 12 Oct 2006