Paper Reading 25 – Finding Your Way in a Multi-dimensional Semantic Space with Luminoso

Comments
Paola Garza
Steven Hennessy

Reference Information
Finding Your Way in a Multi-dimensional Semantic Space with Luminoso
 - Robert Speer, Catherine Havasi, Nichole Treadway, and Henry Lieberman
 - IUI 2010, Hong Kong, China

Summary
The authors of this paper presented a program called Luminoso.  They explained how massive amount of semantic data, such as that from a mass survey, can be very difficult to sort through and form conclusions from.  This is especially true when user free response sections are available for the user to fill in.  What they propose is forming relationships between all of this data and representing it in a multi-dimensional space.  This is exactly what they accomplished with Luminoso.

The main view of the Luminoso interface.

The described the program as appearing similarly to Google Maps in how the user can interact with it.  Each node stands for a particular topic or area, and the more elements that relate to this area, the large the node becomes.  Then related areas branch off of this, and so on.  These clusters form abstract representation in the space.  One of the key features they included was the ability for the user to drag individual nodes in order to realign the focus of the overall arrangement of data.  This allows the user to concentrate on a specific opinion type in a survey for example and see all the various elements relating to it.  The designers argue that using a system like Luminoso to visualize data will become increasingly important as semantic networks continue to grow.

Discussion
I’ll be honest and say some of the content of this paper went a little over my head.  But I do understand the general concept of it all.  One thing I learned from this is that writing anything in the free response sections of a survey is apparently pointless when the survey audience is extremely large.  This is probably obvious to most though.  If no one is going to read it, there really isn’t any point to writing anything in those sections.  However, they mentioned they could group these free response areas with the applicable clusters of data making it easier to view on a handful at a time.  I have never thought of any major needs for a system like Luminoso, but they have convinced me that it can be very useful when dealing with massive amounts of polled or collected data.