Charting Linguistic Features by Density Estimation

点击翻译按钮获取中文摘要

ABSTRACT During the 1980s statistical techniques for analysis for geographical patterns have been refined to the point that they may be applied to data from many different fields. This project will implement density-estimation algorithms to draw differential- frequency boundaries for language data, using the extensive computer files of the Linguistic Atlas of the Middle and South Atlantic States (LAMSAS); these boundaries will then be compared to dialect feature boundaries generated with pre-statistical qualitative methods and to boundaries of selected topographical and cultural areas. Prior NSF support assisted the development of efficient small-computer methods to encode LAMSAS data and to develop preliminary methods for its statistical analysis. This project will process files from the existing data-bases with statistical algorithms and with GIS (Geographical Information System) software for displays; a Sun work station will provide fast processing for computer-intensive statistics and graphics. Specific programs will be written by a consultant to implement density- estimation for LAMSAS data and to enable comparison with digitized maps of older boundaries under GIS. Approximately 50 items will receive comprehensive analysis and comparison, leading to several articles on the nature of areal variation when it is conceived as the differential frequency of linguistic features, and on the relation of this quantitative method to earlier qualitative methods.