One month ago (already !) was Carto2.0, a conference organized by 3 actors of the French Visual Thinking scene :

• Francis Moaty : teacher at ESIEE, co-founder and director of the MISTE (Master in Scientific, Technical and Economical Intelligence)
• Claude Aschenbrenner : entreprise architect at Euroclear, blogger and member of the pedagogical team at MISTE
• Christophe Tricot : PhD, research engineer at Mondeca and blogger

The papers presented at that conference can be downloaded here, or viewed here (in Flash format), but since everything was in French, I am going to report a few bits in English as I believe it could interest more people than only French speaking readers

I was invited to present some of the research I had done for a previous blog entry on navigation interfaces for information sites last November (see a French version here : Les interfaces de navigation pour les sites d'information ). That included some best practices from leading French sites such as Rue89.com, Agoravox, … (see Social Media Club Paris for more info), but let's speak about what I have seen.

I have missed the opening words from Alain Juillet, the French Chief Representative for Economic Intelligence at SGDN, but was enthralled by Guy Melançon's opening plenary session. A doctor in mathematics from UQAM (Montréal), he now does research for CNRS' LaBRI and INRIA's Futurs - Gravite labs in Bordeaux. He has a background in graph drawing and has worked on pattern recognition in graphs. He mentioned a few very interesting references for people dealing with data-mining, BI and in general heavy information interface design :

• Visual Analytics (Thomas & Cook, 2006)
• Multi-faceted insight through interoperable images (Alfred Inselberg)
• KDD panel, the perfect data-mining tool : automated or interactive (Mihael Ankerst, 2002)
• Zoom, filter, analyze : the "Details-on-Demand" approach (Daniel Keim, 2006)
• and of course "The eyes have it" (Ben Shneiderman, 1996)
  

And he offered a solution to Tufte's limits adapted to Moore's law : make the maps interactive

He is also a strong advocate of crowdsourcing as a collaborative approach to looking for patterns in massive databases of data. He mentioned the example of IBM's Many Eyes experiment, where viewer's comments will reveal phenomenons and create new knowledge (see Catherine Lenglet's article on this blog last year)

We then had a presentation by Serge Guillard on business mapping. A director at Mark Company, he is a strong advocate of organization mapping (process, org charts, activities, procedures, job descriptions, …). His presentation was very exhaustive, including examples of work he'd done for Carglass and the CNES. Serge is the co-author of a book on graphical process description : free Visio templates to download here.

We then had very interesting presentations from leading French companies in the data crawling / intelligent agents fields (Cybion) and search engine / data visualization fields (Kartoo). 

Cybion was started in 1995 by Carlo Revelli (also the founder of Agoravox), and quickly made a name in search, with some of the first intelligent agents, crawling the web to gather information on a given subject (we did not have Google at that time ;-). Their presentation (made by Mounir Rochdi, founder of the first competitive intelligence community in Marocco) was on risk analysis and crisis anticipation on internet : identification, mapping, evaluation, simulation and display on a dashboard.

The next presentation was by Laurent Baleydier, CEO and founder of Kartoo (founded in 2001, see a recent article by The Next Web). He presented 3 strategies for search on Internet, very smart approaches that can help us sorting out information :

• take a bird's eye view, using semantic maps, and then zoom in on the zones of interest (Kartoo's speciality but also Quintura) ;
• start with a fixed / known point and navigate away from it (think Musicovery, Flickr of YouTube network graphs) ;
• or for those complex searches (10% of the requests = 50% of the time spent searching), use tools that recommend keywords to refine search such as Grokker or KVisu (Kartoo's new tool).

As an interesting feedback, he shared with us that most of KVisu's search users start with the list and after 1 months usage of the maps increase by 30%. This proves a rather quick learning curve that was made possible through a very structured usability  / HCI design process of the tool. Kartoo is a Google partner (they distribute their Google Mini and Search Appliance). See more info on their blog.

The last presentation I attended was Maxime Crepel and Dominique Boullier's. Dominique, whom I first met as head of the Lutin usability lab, presented us with Maxime their research objectives. They focus on tagclouds as graphical interface for navigation (and you know I am a fan ;-). Also the director of the Las lab (Sociology and Anthropology) in Rennes (see the blog he co-author with Audrey Lohard on virtual worlds as part of a research project), he works with Maxime (PhD student, also engineer at the SENSE lab at Orange R&D), whom I had already heard as he worked on "weak cooperations" with Dominique Cardon and Nicolas Pissard (now with faberNovel), and some other projects (see the website anoptique.org).

They are among other things trying to measure the level of information depth (colors, font size) that is appropriate for an optimal use.

Among the presentations I have missed (but I had the opportunity to chat with them and re-discover their work) were (non exhaustive list) :

• Guilhem Fouetillou's. He is the co-founder and President of the RTGI Group, the editor of a very nice visualization engine that was used during the last French presidential election to display the Blogopole, a map of all political blogs by affinity. See a very interesting presentation on Slideshare describing all kinds of social network analysis he did. He is now doing it for the US 2008 election :

• Jean Delahousse, founder of Mondeca, a French semantic web company, who did a presentation of cartography and ontologies.

• Olivier Nerot, founder of amoWeba and Social Computing, who designed the network navigation tool used by Societe.com to analyze company's board members, and investors links with other individuals or companies. A great competitive intelligence tool :

• Christophe Douy, from Pikko Software, editor of the Arak4Wiki solution, a complete solution for wiki editing and navigation, including EasyKube (treemaps), VisionLink (network graphs), etc… that looks very powerful and builds on works from the same CNRS Labri lab from Bordeaux where Guy Melancon works.

• Leave a comment... (0)

A few weeks ago was held the second meeting of the Social Media Club Paris, started by Pierre-Yves P. (one of my 2 partners at faberNovel Consulting) and Alban M. (renowned author of "L'Age de Peer" and currently working for Orange - France Telecom). We had around the table the founders of some of the most advanced participative news sites in France : L. Mauriac (Rue89), C. Revelli (Agoravox), B. Thieulin (Desirsdavenir), G. Champeau (Ratiatum), J. Jacob (Obiwi), etc… Among the very interesting things that were said that evening on fostering and managing online participation and debates, there were a few on visualization.

Technology, said T. Vedel (IEP Paris), brings some solutions and create new difficulties. Two avenues are possible:

• aggregation or large quantities of information with visualization tools (mapping, semantic analysis, synopti vision, …) to facilitate the entry in the debate
• or the use of knowledge trees (see below the example done synthesizing motions of the 2005 Congress of the French Party Socialiste), using color codes to show votes in a more efficient manner than arithmetic averages

The problem with dealing with a large quantity of information he said is our "limited cognitive abilities and cultural resources". "The image is the dominant language in politics as it has the capacity both to transport a lot of information and emotion" he concluded.

The issues that have been experimented in the political science field (where debates are a core component) are also critical to modern information websites, where a debate/discussion happens in the comments and extend the article, but also the mass of articles themselves could be seen as a debate on some topics, showing different arguments and their evolution. I am going to list the most used here, explain their limits and illustrate them with examples

• Tagcloud: I have already presented it as it is one of my favorite knowledge design techniques, because I think it is both simple and has great potential for KM. Originally designed to display keywords associated to a text by the author and readers, but also experimentally used to synthesize, or present key arguments in a text (see the analysis of French candidates political speaches by French linguist Jean Veronis). You can read more at Peter Vander Wal's Infocloud: he is the one who coined the term "folksonomy"). A free tool to create tagclouds is TagCrowd. Below is an example taken from Amazon's new Inside This Book concordance analysis feature, available on books that are part of the Search Inside the Book program.

As you can judge, this feature is not yet very useful : can you guess what the "Tipping Point" book from Malcolm Gladwell is all about ? Not really a synthesis You see that to be useful you have to trim your text from non-significant or not-significant-enough words, and probably go for a different algorythm altogether than the most frequent words.

Amazon is experimenting with various (and numerous !) techniques to synthesize texts but none really seem very effective: Key Phrases (Statistically Improbable Phrases, or Capitalized Phrases) don't return some of the words or phrases you would expect such as mavens, connectors, … Text Stats will only interest fans of bizarre statistics or adepts of Freakonomics. Only the Citations give you some meaningful data, but not really useful the way they are presented here. But I will come back to this later (with the Touchgraph tool).

There are 2 other ones you can find : heatmaps (applied on treemaps) and network maps.

• Heatmap: one of the best example can be found at Newsmap, presenting live news from Google News (via Infosthetics), used on Rue89.com, both using the Marumushi tool

Another nice heatmap is this one (via Smashing Magazine), displaying the zones of intense debate/battle on Wikipedia.

• Networkmaps: very much used for music discovery, LivePlasma's engine has been used at CNet News.com in 2005 under the name BigPicture, but this flash widget has disappeared recently (see Robert Scoble's post and a more recent one here with a link to a similar application by LinkSViewer). One of the best apps for this is TouchGraph, which has been adapted for Amazon (with books often bought by similar profiles), Google (with links between websites), Facebook (with pictures from friends) or Del.icio.us (with related tags).

So ? Some taste of the semantic web ? So far only the heatmap seems to have lasting success with information sites, the network map didn't stick so far, despite good reviews from experts (see Olivier Ertzscheid's blog). Those new interfaces still have to be improved a lot to be used by casual, everyday users, but are already very useful for some professionals in the information analysis sector, which seems to be the primary market for the editors of those tools. Adam Ostrow at Mashable has a good selection of dataviz tools you might want to check, as well as the Smashing Magazine.

We have also seen some original / innovative visualization techniques popping up here and there, some taking a social or time approach to linking and visualizing content:

• Radial graphs: some are used to display social connections in a network or community around an individual, like those influenced by Facebook's Friend Wheel and the idea of a "social graph" representing all connections between members of a community in a graphical way. For others, the idea is to represent the 4th dimension of time as a wheel, and adding the connections inside it like spokes.

But again, also it may look nice, understanding it is not intuitive, and time is still best shown … as time like in many animated graphics (see the best example with Hans Rosling's Trendanalyzer) or using a timeline (see Dandelife).

The last solutions are using 3D environments, either pure mathematical constructions such as network representations, very similar in a way to space/galaxy navigation see also the incredible Universe experiment by J. Harris, or real-life environments, based for example on city or landscape metaphors. More advanced metaverse metaphors could be developped, using diverse immersive tools, like those displayed in movies (eg Disclosure, starring Michael Douglas, and Demi Moore) and sci-fi books (eg Tom Clancy's NetForce collection of books) or the "must" in VR machines such as the "Virtusphere" shown below.

Let's hope someone will come up with better stuff than that to explore information

• Leave a comment... (1)