También disponible en Español


The digital magazine of InfoVis.net

by Juan C. Dürsteler [message nº 151]

Finding information hidden in huge amounts of data is one of the constant leitmotivs of Information Visualisation. Autofocus is a software that combines computational linguistics with visualisation to achieve this goal. 
ClusterMap.gif (53328 bytes)
ClusterMap in AutoFocus
Screenshot of the author's computer.
Click on the image to enlarge it.

Autofocus is a visualisation software that has recently appeared in the marketplace in the hands of Aduna, a transmutation of the old Aidministrator (see number 62), a Dutch company devoted to the creation of open software that takes advantage of the semantic web and of its visualisation.

The goal of this program is to simplify the search for existing files in our computer, e-mail or websites through the advanced use of elements pertaining to the domains of visualisation and information retrieval.

For the sake of evaluation I have downloaded the trial version (30 days) and I have installed it on my computer.

Once installed and running the first thing you have to do is to define an information source and scan it. I’ve begun with InfoVis.net. The scan took nearly 46 minutes through a 512 Kb/s ADSL connection. After that I’ve scanned my personal directory on the hard disk (6.9 GB over 36,685 files) in about 26 minutes. Autofocus uses this time obviously to build indexes and statistical information that allows very quick further retrievals.

Once scanned, the program has generated its indexes and statistics so that you can begin to query it receiving very quick responses. 

On the left margin of the screen there is a text input area where you can introduce the search terms. To its right you immediately get the results of the query, be them files, e-mails or web pages.

Below the text input you find two windows that allow you to see a historical archive of the search terms already used and the suggestions that the software provides within the framework of what Aduna calls Guided Exploration.

Autofocus1.gif (104949 bytes) Autofocus2.gif (91576 bytes)
AutoFocus showing two of the possible configurations: In both of them you can appreciate above to the left the dialogue to enter the search terms.  Below you find two windows: one with the historic archive of the terms already entered for the search and the suggestions of new terms that the program provides. To the right you find the cluster map and below it the retrieved information . This can be presented as lines of text (like in MS Explorer in the left image) or with short annotations resembling the style of traditional search engines  (á la Google, in the right image).
The icons that appear on the right margin of the screen allow you to change the colour scheme configuration and the views of the map.
 Source: Screenshot of the author's computer.
Click on the images to enlarge them.

In its widest sense Guided Exploration means that the system proposes you views, elements or procedures that help (guide) you in the search. In this particular case it means that Autofocus finds a series of words statistically related with the ones we are looking for and offers them as suggestions. If we choose one of them it performs a new search including this word and in the graphic you can find the new documents that fit the query along with their relationship with the others already present. The objective of all this is to help us refine the search.

The results are presented both in textual form, as classical hyperlinks that lead us to the “clicked” files, and in graphical form using an interactive cluster map that unfolds elegantly showing the results of the query within ovals that enclose other smaller circles. 

Different groups of circles can be linked so that the enclosing oval covers all of them narrowing the gaps in between (see the figures above) symbolising the relations existing between groups. This allows us to get an overview of the search results (the clusters) which we can interact with, by moving them, erasing them (deleting results of our search) or clicking on them to open the files or directories they refer to.

Although it’s easy to understand, the interpretation of a Cluster Map is not evident. Basically every result cluster connects to one or more groups by means of one or more of the search terms used. The group that is connected to all the terms supplied is the set of documents that best represent what we are looking for.

Autofocus is available in two versions Personal and Server. The first acts as a personal use program whereas the server version can act as a search engine for an Intranet or a web site.

As happens with Grokker, KarTOO or Vivisimo (see number 138) Autofocus tries to fill the gap that separates us from the bulk of information available, be it on our hard disk or in Internet, by using sophisticated techniques and advanced visualisations. Every one of them has a slightly different philosophy. 

On the pros side of Autofocus you find the impressive speed at which it retrieves results, the guided search and the elegance of the visualisation. Also in favour is the versatility of types of format that it can read and treat  

Format ASCII  Web  Acrobat  MS Office MS Works Open Office  Wordperfect Rich Text Format 
Extensions *.txt *.html, *.xml *.pdf *.doc, *.xls, *.ppt *.wks *.sx* *.wpd ... *.rtf

On the cons side is the need to do a scan that can last for a long time and also that we aren’t notified that it has been updated automatically.

Grokker is somewhat slower but you don’t need to perform a previous scan of the source (at least when searchin in Internet). Its visualisation is, in my opinion, more intuitive and easier to understand without any further explanation.

The search for ways to find the information we need in a more intuitive way, using the power of the computational linguistics combined with visualisation has another exponent in Autofocus, backed up by the experience of Aduna.

Nevertheless I think that in all the above-mentioned cases we haven’t yet found the really intuitive visualisation for the normal, graphics and visualisation illiterate, user for whom we need to create self explaining and self evident systems or it will be very difficult to bridge the gap between the traditional, textual-output search engines and the ones that will let us quickly find what we are looking for in a clear and intuitive way.

I would like to thank Christiaan Fluit from Aduna for his interesting reflections about Autofocus and his help with AutoFocus.

Links of this issue:

http://aduna.biz   Website de ADUNA
http://www.infovis.net/printMag.php?num=62&lang=2   Number 62 Visualising the Semantic Web
http://www.infovis.net/printMag.php?num=26&lang=2   Number 26 about Semantic Web
http://aduna.biz/products/autofocus/personal/download.html   Download page for Autofocus Personal
http://www.infovis.net/printMag.php?num=138&lang=2   Number 138 about Grokker, or the Visual Navigation
© Copyright InfoVis.net 2000-2014