(2 of 3)
In the rich, fruity tones of the classroom vaudevillian or the dusty mumble of the archives burrower, scholars at the Dartmouth conference "interface" on such matters as "Semantic Networks, Frames and a Conventional Data Base Methodology" and "Controlled Random Generated Parameter Definition." At any given time, several papers are being delivered. What a wanderer discovers depends on which classroom door he opens. It also depends on how familiar he becomes with such terms as COBOL, SAS and FORTRAN (computer languages) and lemmatization (grouping together the variant forms of the same word). Not to mention all-purpose academic sentence deadeners like parameters (a specific statistical term used loosely to mean limits).
Accounts fly about of such worthy and weighty computer projects as the compiling of vast concordances and the storage and retrieval of texts (Latin and Greek at Dartmouth, 17th and 18th century English at Cleveland State University). Machine translation projects may be emerging from a decade of disrepute brought on by unrealistically high expectations that foundered on the marvelous complexities of language. No one in the corridors at Dartmouth thinks that literary translations will ever be done by computer— War and Peace in Russian fed into one end of the machine with a readable English version emerging at the other. But machine translation may work with technical prose of sharply limited vocabulary.
Walk through the right door and there is Peter J. McGuire of Georgia Tech moving determinedly "Toward the Development of an Algorithm Permitting Computer Evaluation of Coherence in Prose." An algorithm is a mathematical scheme followed to solve a problem, not so? Very well, forward march: McGuire, a young man who teaches English to technicians, had noticed a kind of student writing that produced in the reader the feeling that he had learned a lot, but could not remember anything. The trouble? Lack of coherence. What helps coherence? A lot of sturdy ordinals—"firsts," "seconds," "thirds—and plenty of vigorous "thus-es." What hurts coherence? Free-floating "this-es" and "these-es" that do not refer back to anything, not to mention phony locutions like "From this, we can see that..." when in fact nothing is in sight. Possible solution? Run a lot of student papers already graded for coherence or lack of it through a computer to count positive and negative elements, compare with readers' reactions, refine and rework the process, and repeat. Result? Cautious optimism, says the coherent and careful McGuire.