Computer-Produced Representation of Dialectal Variation: Initial Fricatives in Southern British English

W. Nelson Francis, Jan Svartvik, Gerald M. RubinCOLING 1969

TL;DR

This paper introduces a novel approach to dialectology by leveraging computer processing to analyze and visualize linguistic variation in Southern British English. By automating the sorting and mapping of dialectal data, the authors demonstrate a significant reduction in manual effort, enabling more comprehensive and nuanced studies of linguistic features.

What Problem Does This Paper Address?

Traditional dialectology, particularly in its interpretive phase, involves the analysis of vast amounts of linguistic data, often in the form of words and short phrases. This data needs to be sorted and compared across various linguistic bases: phonological, morphological, lexical, and geographical. The primary obstacle to widespread computer use in this field has been the formidable task of preparing historical dialect survey data, often recorded in narrow phonetic transcriptions, for computational analysis. This manual preparation is time-consuming and prone to error, hindering the exploration of complex dialectal patterns.

Methods

The authors propose a computer-driven methodology for analyzing dialectal variation. The core of their approach involves:

  • Data Preparation: Transcribing dialectal data into a standardized, computer-readable format. This involved selecting specific words (initial fricatives in Southern British English) and recording their phonetic forms from various localities.
  • Computerized Sorting: Utilizing computer programs to sort the prepared data in multiple ways. This included sorting by keyword, locality, and phonetic transcription, generating several lists (List I-IV) that facilitated initial examination.
  • Computerized Mapping: Employing a CalComp plotter to generate dialect maps. This involved writing FORTRAN subroutines to instruct the plotter to draw map outlines and place phonetic symbols at specific locality coordinates. The process involved digitizing map outlines and processing linguistic data into a format suitable for the plotter.

The novelty lies in the systematic application of computer techniques to overcome the manual labor inherent in traditional dialect mapping, allowing for more intricate analyses and visualizations.

Datasets

The primary dataset used in this study is derived from the Survey of English Dialects (SED). Specifically, the authors utilized data from Volume 4, which covers ten southeastern counties of England. The data selected for examination focused on initial fricatives (f-, s-, th-) in a list of 68 words. For their analysis, they used data from 75 localities within this region.

Key Insights and Ideas

  • Efficiency Gains: Computerized sorting and mapping can achieve in hours what would take endless hours by traditional methods, significantly accelerating dialectological research.
  • Granularity of Variation: The study revealed a remarkable degree of variation, not only between localities but also on a word-by-word basis, supporting the idea that "Every word has its own history."
  • Influence of Following Sounds: The research suggests that the sounds immediately following an initial fricative can influence its voicing, a complex interaction that is difficult to untangle with manual methods.
  • Geographical Patterns: The maps generated clearly delineate isoglosses (lines of linguistic division), highlighting sharp boundaries in voicing patterns, particularly on the eastern edge of the voicing area.
  • Challenging Traditional Assumptions: The distribution of voicing, especially the eastern extension of voicing in certain words, presented a puzzle that challenged existing historical views of language spread.

Key Results

The study successfully demonstrated the feasibility and advantages of computer-assisted dialectology. Key results include:

  • Data Processing: The preparation of 5100 data items (68 words x 75 localities) for computer analysis.
  • Automated Output: The generation of four sets of sorted lists and several types of maps, including individual word maps and proportional maps showing voiceless:voiced ratios for different word sets.
  • Detailed Variation: For instance, the f-words showed a voiceless:voiced ratio ranging from 20:54 to 52:22, illustrating significant variation.
  • Geographical Delineation: The maps clearly showed distinct areas of high voicing, mixed areas, and fringe areas with limited voicing, with a particularly strong isogloss on the eastern boundary.
  • Word-Specific Patterns: The study highlighted that even words within the same phonetic category (e.g., f-words) exhibited unique distribution patterns.

Limitations and Future Work

The authors acknowledge that their study is a preliminary exploration and that further in-depth analysis is required.

  • Further Study Needed: More detailed study of individual words and localities is necessary to fully unravel the complexities of dialectal features like initial fricative voicing.
  • Incidental Conversation Data: The authors suggest that incorporating data from incidental conversation, in addition to citation forms, could reveal a higher incidence of initial voicing, as these forms are often more natural.
  • Exploration of Implications: The full linguistic implications of the observed distributions, particularly the puzzling eastern extension of voicing, are deferred to a later study.
  • Broader Application: The authors express hope that their work demonstrates that with adequate and convenient data, computers can be an invaluable aid to dialectologists, opening new avenues for research in linguistic geography.

Summary updated

BibTeX
@inproceedings{francis1969computer,
  author = {W. Nelson Francis and Jan Svartvik and Gerald M. Rubin},
  title = {{Computer-Produced Representation of Dialectal Variation: Initial Fricatives in Southern British English}},
  booktitle = {International Conference on Computational Linguistics},
  year = {1969},
}
Figure