Hawk - Experimental Tool for Information Visualization

download Hawk-V.tgz    view screendumps

About Hawk
Hawk visualizes information structures in database systems and other data sources. By defining semantic variables (much like Fuzzy Logic) based on information from an EOF adaptor, sub-concept/super-concept relationships are being computed by using methods from the mathematical discipline of conceptual analysis. Data is visualized as a directed acyclic graph, the graphical representation supports "pruning operations" for better handling.

Quick Start
You can start by downloading the Hawk software from this page. Please unpack the software using Opener.app, or by typing gnutar xfz Hawk-V.tgz at the command line. Since Hawk makes use of EOF, you need a populated database. In order to get you started quickly, I have included an EOModel which works with OpenBase Light (part of the MacOS X Server distribution) by using the Movies example.

After starting Hawk.app, open the file movies.hfscale inside the app wrapper. Hawk will load a sample conceptual scale (a set of semantic variables and their definition) from the file, together with an EOModel stored in the app wrapper. The database is then being queried for documents matching these variables, and the result is displayed as a directed Graph.

The following page has a guided tour of Hawk, together with some screen dumps.

Why not just use Sherlock?
Sherlock does an excellent job of implementing fast, multithreaded searches. However, like most web search facilities such as AltaVista or WebCrawler, search results are displayed as a ranked, linear list. This has the drawback that any internal relationships of the search results such as similarity and cardinality of search term results are invisible.

If, for example, you search for search terms A, B, and C, documents with one or more of the search terms are listed, usually ranked by confidence. What you do not get to see is how many of these documents contain A and B as opposed to those which contain A only. Likewise, it may be of interest whether for example search term C yields few results or none at all, and may have to be relaxed somewhat.

By displaying such groups of documents with common properties as a node in a Graph, underlying structures in search results can be visualized.

Thus, Hawk is not a replacement for search technologies such as Sherlock, but can be used as an additional visual interface for browsing results.

Future Directions
Hawk is work in progress, and you are looking at a very early stage which still contains many bugs, and has a somewhat awkward user interface. Most of all, Hawk needs a better tool for editing conceptual scales. I am leaning towards deploying the Fetch Specification Editor built into the new EOModeler, but Brian Hill's EOQuery also looks quite promising.

Of course, in order to be compatible with MacOS X, Hawk will need to be ported to the Quartz libraries.

Current research is on developing a concept based Focus+Context mechanism the will help reducing the complexity of large graphs, which are currently difficult to use. I also intend to experiment with 3-dimensional Visualization Metaphors such as the Information Cone Tree developped at Xerox PARC.


Christian Neuss, November 1999