Lay it out an an Xchars, the way I do Huffman trees, except the weighting graph is a two dimensional spread of cells. The layout of the weightings, their position on the graph is scaled, and alost all grammars will be sphere packinsg, they will work best over their domain when the weightings re spread to resemble the balance Hffan tree, overlayed. Up the Y axis ae te heavy weights, and they are rarer, the bottom elements have varying amounts of left/right orderings as they are equally probable horizontially.
This is a classic case for a Deep mind, looking at myriads of text, finding its own meaning. Shunting algorithm, needs study, for AI, really.
No comments:
Post a Comment