Use of an interactive program in analyzing data for a dialect dictionary

Abstract

THE Dictionary of American Regional English (DARE) will be the official dictionary of the American Dialect Society, which was founded in 1889 in the hope of someday producing a dictionary which would be the American equivalent of Joseph Wright's English Dialect Dictionary. Not until recent years has it been possible to collect and process enough data to produce DARE, but during the past eighty-six years collection has been going on with a view to eventual publication. Now that funds and computers are available, we expect within a very few years to have the dictionary published by Harvard University's Belknap Press. Some examples of the sorts of entries to appear in the dictionary will characterize its distinctness from more conventional dictionaries of standard lexicons. Having searched through a large body of written material and written collections of oral data, we have collected a huge corpus of contemporary speech: words, expressions, grammatical forms, and pronunciations which are local or regional, rather than general, in the United States. Some words which have less than national currency, for example, are rack for a coat hanger in North Carolina, chunk meaning "to throw" in South Carolina, and beau-dollar for a silver dollar, exclusively in the southeast. For a place in the mountains or hills where it is possible to get through without climbing over the top, many people in the northeast who don't say pass say notch; and people in the west who don't say pass say canyon. Pass, the general word, will not be included in the dictionary. Our written data was typed, machine-scanned, and sorted into an alphabetical listing. Our oral data, because of its size, has been treated somewhat differently. In 1002 communities, spread across all 50 stat s and distributed according to the density of the settled population, our fieldworkers interviewed 2752 informants. These people varied in age, sex, race, education, and community type (the distribution of these is shown as Figure 1), but what they had in common was that they were native speakers of English who had lived virtually all of their lives in the areas where we interviewed them. In each community a long Questionair of 1847 questions was filled in by the fieldworkers, who took down the informants' responses to each question (most of them of the form "What do you call such-and-such around here?") with notes and phonetic transcriptions where necessary, for a total of about 2,500,000 responses. Two and a half million responses with an average length of about 20 characters, with all the informant's data tied to each one, will not fit economically in a computer file. We calculated that if we included everything we would need to store 65 million characters. What we have done is to store the words on paper, where we human editors can get at them and make editorial judgments about them, and to store the numbers in a computer file, where the machine can get at them and count them in various ways. The kinds of questions we asked for each response are (a) is it general or limited; (b) if it is not general, does it vary regionally or socially; (c) in what respects does it vary; and (d). how does its distribution compare with the distribution of other responses? Answers to these questions are found on a printout of all the responses to a given question. Two of these appear as Figures 2 and 3. We see that sunrise and sunset are the most