Minutes Guided Tour, September 6th, 2006

Notes

We arranged a tour with a tour guide at the Rijksmuseum Philips Wing to make observations that could apply to making the CHIP personalized tour. The highly recommended tour guide Bregtje Viergever gave us a personally and professionally interesting tour of the Philips Wing. She gave for the most part a "normal" tour, interacting with us as she normally would, but was also aware of what we wanted to learn from it.

The following subsections describe technique Bregtje uses and how they could apply to CHIP.

Getting Started

Initial User Assessment Bregtje often figures out in the beginning of the tour what type of group it is, and then makes choices from the start about how to form the tour. For example, if the group is from a ship building company, then she will show all the pieces that have to do with building ships. One equivalent in the CHIP personalized tour would be to look at the user profile. Given the current demo, this would be looking at the user's interests and selecting artwork that suits these interest. Another equivalent would be letting the user specify a topic at the start of the tour for the tour to follow.

Addressing Expectations Early Bregtje most often starts her tours with "we will see The Nightwatch at the end of the tour". Every visitor wants to see it, so it is important to reassure them that they will see it, and when they can expect it so tension doesn't build (not negatively, anyway). This statement is also important for her tours because she often builds several themes that culminate in seeing "The Nightwatch", thus the anticipation of seeing the most famous work synchronizes with narrative build up toward that work. In CHIP, we could indicate at the beginning what the largest theme is and how (and when) it will end. More specifically, we could select one work, such as "The Nightwatch", or perhaps some particularly important work for the given user, and make the equivalent announcement at the start, and select aspects of that work to build one or more continuing themes around while going through other artworks while approaching the final one.

Story Templates Bregtje has several set stories or progressions along certain artworks that she keeps in her head. Often she sticks to one such template, which keeps the general story and its narrative intact, but adjusts the details to the audience as the tour progresses. The CHIP tour program could have such templates. The large scale presentation remains essentially the same, but CHIP automatically adjusts the details based on user interaction. CHIP could also choose one of multiple templates based on the particular user.

Moving Along

Continuing Themes Every story or template as several continuing themes. These are topics that occur in multiple artworks in the tour, for which each artwork can add more to understanding that topic. For example, Bregtje frequently showed us paintings with intentionally thick brush strokes, which, when covered in varnish, added a real highlight over the shiny bump of varnish, which then adds to the underlying image. The CHIP personalized tour could select or detect repeated topics that occur among the artworks it selects to present to the user. When showing each artwork, the tour could be sure to mention these topics within the artwork, or it could emphasize these topics within the artwork display by, for example, putting it first or on top, or adding text that clarifies that the topic is important and how it fits within the progression of its continuing theme in the tour.

Ongoing User Observation While giving the tour, Bregtje observes how the user responds to what she says and to the artworks themselves. The primary cues are lack of interest, when the users are non interacting or otherwise clearly bored, and focussed interest, when the users ask about particular topics or look closely at details of the artwork. CHIP could detect lack of interest by the user clicking quickly away to other topics (although this is imprecise in web interaction). CHIP could observe focussed interest by what topics user click on. If the demo can observe zooming in and out of large artwork images, then CHIP could associated focussed upon details with annotated topics and note that the user is interested in these topics.

Observation Responses Bregte responds to lack of interest by moving along immediately to the next artwork. The CHIP equivalent would be to show less information for the artwork and to facilitate letting the user move onward. One way in which Bregtje responds to focussed interest by noting what topic the user showed interest in and making it a continuing theme to use in the rest of the tour. CHIP could also add such detected topics to the themes to follow in the rest of the generated tour. Another technique of Bregtje, and potentially for CHIP, is to select artworks to add to the rest of the tour that illustate these new continuing themes well.

Projected End Bregtje plans to end the tour at a particular time. She uses some of the techniques described below to make sure it does. CHIP could similarly measure a tour either in terms of time or number of artworks and topics. CHIP could also calculate at the start and at points during the tour what the projected end is and then adjust the plan for the rest of the tour so that the projected end matches the desired end.

Particular Techniques

Selective Removal When a tour is taking too long, Bregtje makes sure the tour will end promptly by selecrively removing artworks from the rest of the tour. Artworks that are least supportive of the continuing themes are the best to remove. CHIP could perform render-time selective removal when starting the tour by taking an initial draft tour, potentially from a template, and removing from it the artworks that support the given tour goals the least until a desired number of artworks or tour length is acheived. CHIP could also also perform dynamic selective removal by detecting when the tour progression is behind schedule and remove least supportive artworks from the rest of the tour until the projected end time is back on target.

Selective Addition On the other hand, when a tour is going more quickly than planned, Bregtje can select additional artworks to show. Equivalently, she, and CHIP, would select the artworks that are the most supportive to add. CHIP could also add the number of additional artworks that bring the projected of the tour back on schedule. Selective addition happens in the context of a template, either default or adapted. That is, it already had selected artworks and a structure around them, and is adding new artworks to this preset selection and structure. In the absence of templates, CHIP could start a tour by selecting artworks for a desire topic and user interest, and then try to build structure around it from scratch. This is the approach of the Topia demo.

Details Artworks often have multiple details that the guide can discuss as seperate topics. For a given artwork, Bregtje selects which details to discuss and not to discuss based on how many she has time for and which are most supportive of the tour goals, such as continuing themes. The ARIA dataset has the equivalent of details in ARIA aspects, which are all but the first topics in the pulldown menu on the Website. These are not yet in the RDF, but when they are, the demo could select which to show (and not) with an artwork by applying selective addition and selective removal, but to aspects of an artwork instead of artworks.

Compare and Contrast A rhetorical technique that Bregtje uses often is comparion and constrast. That is, she will take two artworks that illustrated different aspects or possibilities for a given topic, and refer to the other artwork when discussion that topic for its partner. The Rembrandt-Caravaggio exhibition used this for every painting. CHIP could apply this by finding among selected artworks topics that repeat by vary somewhat in detail. It could then find particularly relevant or strong variations of the same topic across two paintings, and be sure to mention one painting when presenting that topic for the other.