The third item we need to learn is the structure of the engineering design data. The structure reveals the composition of the product in terms of the different accumulations of attributes and functions that define it. I select Bayesian belief networks for two reasons: 1) they explicitly express cause and effect; 2) they convey some intuitive understanding of how evidence is propagated. By tracing the propagation of evidence, the engineer can learn the dependencies and independencies of the design elements. The general method for constructing belief networks is to draw arcs from causal nodes to effect nodes and then attach a probability to that arc. The hypothesis I use for cause and effect is based upon the conjecture that seeing lower TRS words with respect to a word that it shares contextual similarity causes us to update our belief that we'll see the higher TRS words. This causal influence is found by pairing words with the highest TRS in the co-occurrence matrix. The strategy for building the network is to link the highest associated words in their own clusters first then to link the words between clusters. Using this algorithm I arrive at the following network. The picture shown is partial - not all nodes and arcs are shown in the interest of legibility.
| BEST Lab | Department of Mechanical Engineering | University of California at Berkeley