H.P. Lovecraft knew one criterium: their archetypical intrinsic unpronouncability. Tolien knew another: mixing for the (English) reader exotic phonemes and word melody (Finnish - Old Norse - Arabic for instance). -- BL
(I remember hearing, years ago, of somebody's theory that all language could be analyzed based on letter patterns using a particular 4 x 7 matrix derived from ancient Arabic or some such occult source. Sounded nutty at the time, especially in its claimed application to free-text information retrieval and relevance-ranking. On the other hand, N-grams do sometimes seem to offer an information-theoretic way to cluster related terms ... and hashing is a well-known and productive approach to some kinds of data manipulation ....) - ^z
Sometimes, new information can be discerned from old data simply by applying a new structure -- any structure! 'Course, there's no way of knowing in advance if this new information is in any way useful or relevant... :)
When in doubt, just XOR together all the characters in an arbitrary result string and discard the first byte. :))
-- Bo Leuf
One of my weird friend Barry's "Laws" is: