Structural Parallelism for Coordination and Ellipsis Resolution

You are reading version 1 of this direction, frozen 2026-10-04. The latest version is version 2.

Version 1 of 227 papers44 conceptsfrozen 2026-10-04

In natural language, we often omit redundant information or link complex ideas using words like and, or, and but. While humans intuitively fill in these gaps, these structures—known as coordination and ellipsis—represent a significant challenge for computational parsers. The core mechanism that allows both humans and machines to resolve these ambiguities is structural parallelism: the principle that linked or elided phrases tend to mirror the syntactic and semantic properties of their neighbors.

The Symmetry of Coordination

Coordination is inherently symmetrical. When we say "the CEO of the company and the manager of the store," the two phrases share a nearly identical internal structure. Rajeev Agarwal and Lois C. Boggess demonstrated in A Simple But Useful Approach to Conjunct Identification that this parallelism is a powerful signal for identifying the boundaries of coordinated phrases, especially in technical text where long-distance dependencies are common.

Standard dependency parsers often struggle with these structures because they rely on local context rather than global symmetry. Yoav Goldberg and Michael Elhadad noted in Inspecting the Structural Biases of Dependency Parsing Algorithms that parsers fail when they cannot capture the "balanced" nature of conjunctions. To fix this, researchers like Deirdre Hogan have proposed generative models that explicitly condition the structure of a second conjunct on the features of the first, effectively using the first half of a sentence as a template for the second (Coordinate Noun Phrase Disambiguation in a Generative Parsing Model).

Reconstructing the Missing: Ellipsis and Gapping

Ellipsis occurs when a part of a sentence is omitted because it is understood from context (e.g., "John ate bread, and Mary [ate] rice"). This "gapping" creates a "missing anchor" problem for parsers. The solution lies in aligning the "remnants" (Mary, rice) with their "correlates" (John, bread) in the preceding clause.

Yoshihide Kato and Shigeki Matsubara showed that these elements consistently share the same grammatical or semantic roles, allowing for high-accuracy reconstruction through role-based matching (Parsing Gapping Constructions Based on Grammatical and Semantic Roles). Similarly, Benoît Sagot and Djamé Seddah developed a "fusion" operation that allows a parser to "borrow" the derivation history of a visible clause to build the hidden structure of an elided one (Modeling and Analysis of Elliptic Coordination by Dynamic Exploitation of Derivation Forests in LTAG Parsing).

Parallelism as a Cognitive and Discourse Constraint

The reliance on parallelism isn't just a parsing trick; it's a fundamental property of how we process language. Andrew Kehler argues that certain elliptical forms are only resolvable when the discourse context triggers a search for structural similarities rather than causal links (Common Topics and Coherent Situations: Interpreting Ellipsis in the Context of Discourse Inference).

This cognitive aspect is further explored in studies of Chinese rhetoric, where structural symmetry acts as a trigger for the brain to distribute attributes across parallel units, seamlessly resolving omissions by treating components as interchangeable (Research on the Meaning Generation Mechanism of Chinese Rhetoric Huwen). By incorporating these linguistically motivated features—such as morphological symmetry or semantic similarity—into modern neural networks, researchers have significantly improved the ability of AI to navigate the complex, "messy" reality of human coordination (Improving a Strong Neural Parser with Conjunction-Specific Features).

Papers in this version

History & citationVersion 2 · 2026-10-04

Current · Permanent link

  1. Version 2 (latest) · 2026-10-04 · 25 papers
  2. Version 1 · 2026-10-04 · 27 papers
Cite this direction

Latest version (a living page):

@misc{lacuna_syntactic_parallelism_535a16ab,
  title = {syntactic parallelism},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/syntactic-parallelism/rdr_535a16abf46f4bb58c561947098b9fe7},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-20},
}

This exact version (version 1, frozen text):

@misc{lacuna_structural_parallelism_for_coordination__535a16ab_v1,
  title = {Structural Parallelism for Coordination and Ellipsis Resolution},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/structural-parallelism-for-coordination-and-ellipsis-resolution/rdr_535a16abf46f4bb58c561947098b9fe7/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-20},
}