(PRWEB) May 23, 2002
Stop and think for a momentÂ If pressed to do so, how would you generate torrents of new potential words that may be easily pronounced and make contact with our implicit concept of the English word? Â It would not be accurate to say that the proverbial ÂtheyÂ can do it. They canÂt and they havenÂt, but Imagination Engines, Inc. (IEI) has.
Over a one month period, an artificial neural network system has built itself into a Âlinguistic brainÂ at the IEI laboratory in St. Louis, MO and autonomously read gigabytes worth of Internet content, thereby capturing the zen of the typical English word. This system was then allowed to daydream new potential functional English words that fit the mold of accepted vocabulary. The result is a collection of over 1 million unique word stems whose number may be rapidly expanded to between 10 and 100 million, through the addition of standard prefixes, suffixes, as well as various case and tense endings. From this listing will inevitably arise thousands of new web domain, product, service, and corporate names.
This linguistic feat was accomplished using IEIÂs patented Creativity MachineÂ® Paradigm (US Patent 5,659,666), a form of synthetic intelligence that has already produced nationally advertised products, composed 11,000 copyrighted melodies, and generated the formulas of nearly one million new potential chemical systems. The power of this inventive technology has now been demonstrated once again, as this newest generation of artificial intelligence has cut an extremely wide swath through all the words that are to beÂ
We note that the only available benchmark to contrast this profound exercise are what are known as Âgenetic algorithms,Â computational systems that emulate the blind and sluggish evolutionary processes that lead to mutation and natural selection of biological species over eons. To use them as functional English word generators, accepted word fragments (i.e., morphemes) must be researched and accumulated by human researchers and then randomly strung together via computer like beads on a necklace. The ÂfitnessÂ of these often nonsensical words are judged by how realistically the resulting letter combinations are at the junctures between such linguistic beads. The inevitable result of this genetic process is that highly nonsensical words are formed that have no commonality or contact with known words. They are simply syllabic jumbles.
In contrast to the genetic programming approach to generating functional English words, new vocabulary is produced in a Creativity Machine via a parallel computation wherein all pieces of the word are mutually aware of each other as the overall word forms. The result is that typically incompatible syllabic units repel each other, while compatible units aggregate together into words that actually make sense. The joint probabilities of different phonetic units closely matches that of accepted words found in English dictionaries. Â That sensible words can be generated via neural networks, and not genetic algorithms should make perfect sense even to those outside the field of AI: language is born in biological neural networks of the brain and not within the genetic apparatus!
Most importantly, however, the IEI Creativity Machines are totally autonomous, reading textual matter, autonomously forming models of the typical word, and then autonomously forming plausible, novel vocabulary. No other form of artificial intelligence technology can make these claims.
On April 18, 2002, over one million new potential words generated by a Creativity Machine were alphabetized and organized into a reference document that has been submitted to the US Library of Congress as a copyrighted literary work. This is a highly unique collection that anticipates the vast majority of new words that will inevitably arise in our language.