A. L. I. C. E. Artificial Intelligence Foundation

Promoting the development and adoption of ALICE and AIML Free Software

A.L.I.C.E. Brain Picture Gallery

All images are © 2001 Dr. Richard S. Wallace and may be reproduced with permission.  Click on an image to see it in isolation.

eyespiral.jpg

eyespiral.jpg

The eye-shaped log spiral plots all 24,000 categories in the ALICE Brain. The spiral itself represents the root. The trees emerging from the root are the patterns recognized by ALICE.  The branching factor for the root is about 2000, but the average branching factor of the second pattern word is only about two.

logspiral.jpg

logspiral.jpg

Those acquainted with my earlier work on logmap sensors and cortical algorithms for visual processing may notice the similarity with the Graphmaster plot. This similarity is in my opinion no coincidence. The same cortical architecture that enables real-time, attention-based visual processing, can in fact be applied to linguistic processing as well.

sraispiral.jpg

sraispiral.jpg

This plot shows 1/4 of the patterns in Srai.aiml, or about 2000 patterns total.  The four big trees in the lower right are all the patterns rooted at the words WHAT, WHEN, WHERE, and WHO respectively (from right to left). These areas show the compression power of the Graphmaster graph at its best.

termtemp.jpg

termtemp.jpg

Plotting the patterns from Atomic.aiml, we are viewing 1/4 of the Graphmaster graph. The yellow circles indicate <terminal> nodes and the magenta are nodes with a <template>.

spiral.jpg

sprial.jpg

The idea is to plot the root of the graph as a spiral, because it has the most branches. With our 24,000 categories, the number of choices for the first word in this plot is exactly 2001 (!) so the spiral has 2001 1st-level branches. The plot shows the graph down to level 3, with words for the first two levels.

spiral2.jpg

spiral2.jpg

The plot shows all of the branches of the graph with 24,000 categories loaded. The spiral itself is the root, with a branching factor of about 2000 [The number of choices for the first word in an input sentence]. The trees show the average branching factor for the second word is around two, although there are a few dense "bushes" representing patterns starting with "I", "YOU", "HOW", "WHAT", "WHEN", "WHERE", "WHO" and a few others.

atomicspiral.jpg

atomicspiral.jpg

The plot shows 1/4 of the categories from the file Atomic.aiml.  The color scheme shows each level of the graph as a different color. Most categories in the ALICE Brain have <that> and <topic> both equal to "*".  The corresponding redundancy in the Graphmaster subgraphs is illustrated in the lower right.

bigspiral.jpg

bigspiral.jpg

This high-resolution 1024x768 plot shows the ALICE Brain with 24,637 categories loaded.  The spine is actually a log spiral, but with an exponent close to unity, so very nearly the same as a linear spiral.  Gray lines indicate nodes with exactly one branch.  Black lines are nodes with two or more branches.  The leaf nodes have two branches because they store both <template> and <filename>.

bigpat.jpg

bigpat.jpg

This "squared spiral" was formed by modulating the log spiral with a sine wave. The resulting figures uses more of the display area than the raw log spiral. The plot also shows the words sorted by word length, which results in a nice balance among the subtree graphs. Large subtrees tend to begin with shorter words, like "A", "I", "YOU", and "WHAT". Longer words tend to have fewer branches. Arranging the graph in this way, the branches near the center of the spiral are generally fewer than the branches near the outside. The picture makes it easier to see the subtrees than, say, an alphabetical ordering, which agglomerates many subtrees at the center.

For more information pleasecontact drwallace@alicebot.org