A Basis for a Formalization of Linguistic Style

Stephen J. GreenACL 1992 · DOI

Introduction

In the field of computational linguistics, "style" has historically been treated as a secondary concern—a surface-level polish applied after the core meaning of a sentence has been established. However, linguistic style is fundamental to how intent is conveyed and how information is processed by a reader. A machine translation or generation system that lacks stylistic awareness may produce grammatically perfect output that nonetheless feels "off" or fails to achieve its rhetorical purpose. This paper addresses the challenge of transforming descriptive, often fuzzy rhetorical concepts into a formal, computable logic by mapping established theories of textual cohesion onto a multi-level stylistic grammar.

The Stylistic Grammar Framework

The proposed framework builds upon the stylistic grammar developed by DiMarco, which organizes stylistic features into a three-level hierarchy:

  1. Primitive Elements: The foundational building blocks of sentence structure and connectivity.
  2. Abstract Elements: Patterns of primitives that form a stylistic "metalanguage."
  3. Stylistic Goals: High-level objectives such as clarity, concreteness, or precision.

The core contribution of this work is the formalization of "Primitive Elements" using the functional linguistic theories of Halliday and Hasan. By focusing on cohesion—the internal ties that hold a text together—the author provides a mathematical basis for stylistic analysis.

Quantifying Cohesion: The Conjunctness Scale

To bridge the gap between theory and computation, the author introduces a numerical scale of "conjunctness" (connectivity). This scale measures the intensity of cohesive ties, specifically focusing on ellipsis—the omission of words that are understood from the surrounding context. By quantifying how much information is elided, the system can assign a score ranging from $conjunct^0$ (neutral) to $conjunct^6$ (highly connective/discordant).

Examples of this scale include:

  • Nominal Ellipsis ($conjunct^3$): "The first expedition was followed by another two [expeditions]."
  • Clausal Ellipsis ($conjunct^6$): "You can borrow my pen if you want [to borrow my pen]."

In this model, ellipsis is not merely a linguistic shortcut; it is a "stylistic dial" that controls reader engagement. The more information that is elided, the more the reader must actively work to recover the missing context, altering the stylistic profile of the sentence.

Stylistic Dynamics: Concord, Discord, and Goals

The framework distinguishes between concordant (normal/predictable) usage and discordant usage. When a sentence employs excessive connectivity, such as clausal ellipsis, it creates a state of "stylistic discord." While discord might seem like a negative trait, the author argues it is a deliberate tool used to achieve specific stylistic goals:

  • Concreteness: Realized through "dissolution" (high discord). By using clausal ellipsis ($conjunct^6$), the writer forces the reader to reconstruct the missing components, thereby emphasizing specific elements and making the text feel more immediate or concrete.
  • Clarity: Realized through "centroschematic" structures. By replacing intense clausal ellipsis with less disruptive verbal ellipsis ($conjunct^4$), the sentence moves from discord back toward concord, resulting in a sense of predictability and precision.

This formalization allows the "unfinished" feeling of certain sentences to be represented mathematically as a shift from concord to discord within the grammar.

Methodology and Proof of Concept

Rather than utilizing modern large-scale datasets, the paper validates its logic using specific linguistic examples from seminal texts, including Halliday and Hasan’s Cohesion in English (1976) and Quirk et al.’s A Comprehensive Grammar of the English Language (1985). These examples serve as a proof of concept, demonstrating that the formal grammar can successfully map linguistic structures to abstract stylistic goals.

Conclusion and Future Directions

The paper successfully demonstrates a roadmap for integrating stylistic control into functional grammars, such as the Nigel grammar used in the Penman generation system. By providing a formal basis for style, the work moves computational linguistics closer to systems that can generate nuanced, goal-oriented text.

However, the author notes several areas for future expansion:

  • Expanding Cohesive Relations: The current model focuses primarily on ellipsis. Future work must integrate other relations such as substitution, reference, and conjunction into the conjunctness scale.
  • Hierarchic Primitives: While this work addresses "connective" primitives (how elements link), a similar formalization is required for "hierarchic" primitives (how elements are nested or ranked).
  • Implementation: At the time of writing, the full implementation of these theories into a functional stylistic analyzer was still in development.

Summary updated

BibTeX
@inproceedings{green1992basis,
  author = {Stephen J. Green},
  title = {{A Basis for a Formalization of Linguistic Style}},
  booktitle = {Annual Meeting of the Association for Computational Linguistics},
  year = {1992},
  doi = {10.3115/981967.982016},
}
Figure