Do LarKC need a Visualization Plugin for Query Results Presentation
(by Yi Zeng)
 |
| A picture of “LarKC” by WICI during the Bled meeting. |
After the Bled meeting,there is a question that I always have in mind : “Do LarKC need a visualization plugin for query results presentation?”.
The Semantic Web vision requires machine accessible and processable [Antoniou, Harmelen 2004], but the query results are for human to judge whether it is good enough for the query task. Specifically in LarKC, if the query and reasoning results are not good enough, sample size needs to be increased and the query cannot be ended.
 |
| David Karger’s layered Cake (From Flicker.com-pshab’s photostream) |
If query results cannot be presented to users in appropriate ways (especially for large scale query results), users may get confused and it might be very hard for them to make decisions. Hence, more time might be needed.
In a Cognitive Psychology study, Jill Larkin and Herber Simon provided evidence on “Why a Diagram is (Sometimes) Worth Ten Thousand Words!”[Larkin, Simon 1987]. Grigoris and Frank also mentioned “Graphs are a powerful tool for human understanding[Antoniou, Harmelen 2004]”.
Hence besides use case specific applications, considering better interactions with users to reduce the query time, do we need to provide general-purpose visualization plugins for result presentation?The quotation from David Karger brings many insights for me : “whatever is in the cake, what people see is the candle!”
February 5th, 2009 at 2:54 pm
Yeah, diagrams are a must for presentations of complex answers.
Here’s a little mashup using zemanta, dbpedia, freebase and graphviz: http://test.infoblow.zemanta.com/infoblow/galaxy/
basic, but still much better than nothing.
bye
Andraz Tori, Zemanta
February 5th, 2009 at 6:44 pm
I agree with Andraz, with one additional twist, that visualisations ought to be designed to conform to an established graphical grammar and be granular, i.e. allow you to see as little or as much as you want.
For an excellent example of how to deal systematically with this kind of issues (in a truly outstanding visualisation environment) I would recommend
Ben Fry’s Visualizing Data: Exploring and Explaining Data with the Processing Environment .
In Europe we have some oustanding talent in this domain, among them Moritz Stefaner from Potsdam.
February 6th, 2009 at 1:19 pm
really the critical question is the “general purpose” part .. its not whether we need to think about result presentation (most certainly!) but whether it makes sense to do so in a general purpose/task independent way! In the actual publication that Karger was presenting when he showed this slide (’The pathetic fallacy of RDF’ - http://swui.semanticweb.org/swui06/papers/Karger/Pathetic_Fallacy.html) you can see that he was very skeptical about a large part of current general purpose visualization tools (those (over-)utilizing the graph metaphor)
February 6th, 2009 at 2:19 pm
To Andraz:
Thank you for the examples. I agree, we need to think about the question especially for representation of complex answers.
To Stefa:
Thank you for the comments, and recommendations. The book is very good and relevant. I like the quotation “Visualizing Data teaches you how to answer questions, not simply display information” from the book very much. Since you mentioned “granular”, yes, in LarKC we do consider knowledge representation and reasoning with multi-level of granularities and multi-views. Some preliminary introduction can be found in http://wiki.larkc.eu/GranularReasoning.
To Valentin:
Special appreciation should be given to you since in this way, we can come to deeper discussion:). I learned a lot through reading your blog, it is great, and I should learn more from you:)
I think there is no doubt that (sometimes) graph can help to understand. Considering general purpose visualization tools, for “Big Fat Graphs” that shows everything, I think we can avoid over-utilization by showing results with multi-level of importance (in a rational manner) with user intervention. If we add limited rationality to visualization, appropriate utilization might be produced. Rational Choice is one of the core spirits of LarKC that brings from human problem solving to the Semantic Web.