Dialectology by Computer

Gordon R. WoodCOLING 1969

TL;DR

This 1969 paper explores how computers can revolutionize the study of American English dialects by enabling more exhaustive analysis of linguistic data. It outlines methods for collecting, transcribing, and analyzing dialectal variations, suggesting that computational approaches can reveal finer-grained distinctions and new insights into regional language patterns.

What Problem Does This Paper Address?

Traditionally, dialectology relied on manual collection and analysis of spoken language, which was time-consuming and limited the scope of investigation. While descriptive linguistics often focused on the linguistic merit of speech, this paper argues for the importance of considering social judgments associated with language use, particularly in urban settings. The challenge is to move beyond simply cataloging spoken patterns to understanding the social dimensions of dialectal variation. The paper aims to demonstrate how computers can overcome the limitations of manual analysis, allowing for more comprehensive examination of linguistic evidence and the introduction of new analytical variables, such as speaker attitudes towards their own language.

Methods

The paper discusses several methods for collecting dialectal data:

  • Controlled Interviews: Investigators ask guided questions and record informant responses phonetically.
  • Printed Questionnaires: Informants circle words they use and leave others unmarked.
  • Pronunciation Passages: Informants read a standard passage aloud.
  • Picture Description: Informants describe numbered pictures, with their responses recorded.

A significant challenge highlighted is transcription for computer analysis. For conventionally spelled words, direct input is straightforward. However, for phonetic transcriptions or tape recordings, researchers must normalize pronunciations to conventional spellings and devise coding systems for sounds not directly represented by the Roman alphabet. The paper suggests using either numerical codes or modified alphabetic codes for phonetic distinctions.

The paper also touches upon grammatical analysis, proposing the use of slot and filler grammars and the establishment of word boundaries for computer processing. It suggests that computer programs can identify patterns in word usage and potentially reveal relationships between different linguistic elements.

Datasets

Not discussed in the paper. The paper focuses on the methodology of using computers for dialectology rather than presenting results from specific, large-scale datasets. It refers to existing studies (e.g., Atwood, Kurath, Wood) and their data collection methods as a basis for discussion.

Key Insights and Ideas

  • Computers as Exhaustive Tools: Computers offer the potential to examine linguistic evidence far more thoroughly than manual methods, allowing for more detailed counts and the introduction of new variables like speaker attitudes.
  • Bridging Descriptive and Social Linguistics: The paper advocates for incorporating social judgments into dialectological analysis, moving beyond purely descriptive accounts to understand the social meaning of linguistic variations.
  • Transcription Challenges: The process of preparing spoken language data for computer analysis, especially phonetic transcriptions, presents significant hurdles requiring normalization and coding strategies.
  • Potential for New Discoveries: Computational analysis can lead to more precise measurements of word occurrence and reveal subtle gradations in word choice across different demographic groups.
  • Beyond Vocabulary: While early computer applications focused on word lists, the paper points towards the potential for analyzing larger units like sentences and exploring grammatical and syntactic variations.
  • Visualizing Data: The paper suggests that computers can be used to create graphical displays of dialectal variations, potentially offering new ways to understand regional language patterns.

Key Results

The paper does not present specific quantitative results from a computer analysis. Instead, it outlines the potential and development of computer procedures in dialectology. It uses examples like the analysis of "faucet" and "spigot" variations (Table I) and phonological variants of "bench" and "nest" (Table II) to illustrate how data could be tabulated and analyzed computationally. The "results" are more about the conceptual framework and the promise of computer-aided dialectology rather than empirical findings from a specific study.

Limitations and Future Work

Limitations: * Transcription Tedium: The paper acknowledges the immense tedium involved in preparing phonetic transcriptions for computer analysis, which can be an "almost insuperable barrier." * Supersegmental Features: The author notes that supersegmental traits like stress and juncture are often omitted from computer analysis due to their complexity and variability. * Grammatical Analysis Complexity: The paper admits that the current "record of variables in syntax is far from clear" and that the analysis of syntax moving from one utterance to the next presents challenges. * Limited Exploitation: The author suggests that dialectologists have not yet fully exploited the capabilities of computers as research tools.

Future Work: * Computer Mapping: Developing computer programs to map sorted dialectal records, potentially revealing regional English maps that differ from traditional ones. * Graphing and Scaling: Utilizing computers to draw graphs and show measured relationships, such as agreement/disagreement on "good" or "bad" English usages between different social groups (e.g., rich vs. poor, young vs. old). * Integrated Search Instruments: Developing programs that can seamlessly search across vocabulary, pronunciation, and sentence structure, acting as an extensive search instrument to identify convergence and divergence of linguistic elements. * Refined Measurement: Continuing to refine measurements of word occurrence and other linguistic features. * Grammar and Syntax: Further development in analyzing regional English grammar and syntax, likely using conventionally spelled transcriptions.

Summary updated

BibTeX
@inproceedings{wood1969dialectology,
  author = {Gordon R. Wood},
  title = {{Dialectology by Computer}},
  booktitle = {International Conference on Computational Linguistics},
  year = {1969},
}
Figure