Present: Duane Searsmith, Loretta Auvil, Vered Goren, Tanya Clement, Pib Burns, Brian Pytlik-Zillig,Stephen Ramsay (Chair).

2:00 Central

Before the meeting started, there was some discussion (and some positive reports) about the newest version of Open Laszlo. We'll try to keep tabs on this, bearing in mind the particular issues members of the Interface Cell had with this framework on the Nora Project.

Martin and Steve are working on a shortlist of specific analytical procedures (in a sense, a more concise version of what's in the "So you want to . . ." document). Since neither Martin nor the list were available for this meeting, it was decided that the list will be wiki-ed and circulated as soon as possible for discussion at next week's meeting.

We discussed the progress of Sara's use case, and reviewed the approach of nora-db and WordHoard. There was some bit of uncertainty about the chunk level we need for Sara's project, so Steve will check with Sara. Loretta is confident that stock NB and SVM itineraries can be used by Bill for this particular investigation. Sara is going to be creating a training corpus using NoraVis (and perhaps the new-and-improved version of NoraVis being worked on the Interface Cell).

A brief discussion of FeatureLens followed, which was actually the occasion for the opening discussion about OL. Loretta is trying to port FeatureLens into OL, so we'll want to discuss that in more detail as it moves along.

We discussed some possibilities for analysis of the Wright collection, and Duane offered some insight into possible methods for analyzing multiple feature sets and combining them (both Loretta and Duane see this a practical possibility for MONK, though the interface might hide the fact that the underlying system is building multiple models). We also discussed some differences between classes determined by the user ("sentimentality", "eroticism") and classes that are already in the metadata (often, the TEI header), such as the date or place of composition. Steve agreed to create a list of the some possible classes in the Wright collection that fall into the latter category.

We ended with some discussion of clustering techniques and the possibility of bringing exploratory classification techniques (like decision trees) to MONK. Several people noted the need to develop visualization techniques that make these kinds of things more appealing to humanist scholars, and Pib noted that the business community has spent a great deal of time developing visualizations of highly quantitative data for people without extensive mathematics backgrounds.

Meeting adjourned at 2:55 Central.