Slide Nine - Learning the Context

This metric score words based on frequency, distribution and rarity. Then, I select only those words with a TRS greater than the mean as those which are descriptive of the design. To learn the context, I generate a matrix measuring the strength of co-occurrence of certified words. This matrix is the fundamental piece of information from which I can learn both the functional generalizations of the design and the decomposition of the design. To learn the generalizations, we would like to cluster words for their shared meaning. To learn the decomposition, I use Bayesian belief networks to learn and express the dependency relationships between words. Let's begin with the clustering.