Catenal Construction: CxG ↔ DG Bridge Type @cite{osborne-gross-2012} #
Defines the CatenalCx type — the fundamental bridge between Construction
Grammar and Dependency Grammar. A catenal construction pairs a CxG
Construction (form–meaning description) with a DG catena witness
(dependency tree + connected node set).
Osborne & Groß (2012) argue that every construction corresponds to a catena.
This type makes that claim checkable: to instantiate CatenalCx, one must
provide a Construction, a DepTree, the relevant node indices, and a
proof that those nodes form a catena.
Key Definitions #
CatenalCx: the bridge structure itselfCatenalCx.singleton: every single word is a trivial catenal constructionCatenalCx.isConstituent: whether the catena is also a constituentCatenalCx.isNonConstituentCatena: the constructions that require catena theory — things that constituency cannot captureCatenalCx.isContiguous: whether the yield is contiguous (non-risen)
Key Results #
CatenalCx.nodes_nonempty: a CatenalCx always has ≥ 1 nodeCatenalCx.singleton_isSingleWord: the singleton constructor yields a single-word constructionconstituent_implies_catena: every constituent is a catena (general statement; per-tree instances verified vianative_decidein the bridge)
Concrete instances and verified claims are in
Phenomena/Constructions/Studies/OsborneGross2012/Data.lean.
A catenal construction (Osborne & Groß 2012): a construction whose form-side corresponds to a catena in a dependency tree. This is the fundamental bridge type connecting CxG and DG — every construction can be witnessed as a catena over some dependency tree.
The construction field carries the CxG description. The tree,
nodes, and catena fields provide the DG verification: the
construction's words form a connected subgraph.
- construction : ConstructionGrammar.Construction
CxG description of the construction
- tree : DepTree
A dependency tree instantiating the construction
Node indices that carry the construction
The construction nodes form a catena
Instances For
Whether the catenal construction is a single word. Single-word constructions are always catenae (trivially connected).
Equations
- cx.isSingleWord = (cx.nodes.length == 1)
Instances For
Whether the catena also forms a constituent (complete subtree rooted at some node). When true, the construction can be captured by constituent structure; when false, the catena concept is essential.
Equations
- cx.isConstituent = DepGrammar.Catena.isConstituent cx.tree.deps cx.tree.words.length cx.nodes
Instances For
Whether the catenal construction is a non-constituent catena — the case that motivates the entire CxG ↔ DG bridge. These constructions cannot be captured by any constituent-based framework: their form-side is connected in dependency structure but does not correspond to any subtree. Idioms ("spill the beans"), LVCs ("take a bath"), and displacement ("beans she spilled") are typical examples.
Equations
Instances For
Whether the catena's yield is contiguous in linear order. A catena with contiguous yield occupies a consecutive span of positions; a non-contiguous catena is a risen catena (@cite{osborne-2019}, Ch 7) — connected in the tree but separated by intervening material in the string.
Equations
- cx.isContiguous = DepGrammar.isInterval (cx.nodes.mergeSort fun (x1 x2 : ℕ) => decide (x1 ≤ x2))
Instances For
Smart constructor: any single word in a tree trivially forms a CatenalCx.
Uses singleton_isCatena — a single node is always connected.
Equations
Instances For
A singleton CatenalCx is always a single word.
A singleton CatenalCx always has node count 1.
A singleton CatenalCx always has contiguous yield (trivially).
Constituent → Catena: every constituent is a catena. Constituents are complete subtrees (projections rooted at some node), and complete subtrees are connected in the dependency tree.
This is the fundamental containment result establishing that catenae strictly generalize constituents: {constituents} ⊂ {catenae} ⊂ {subsets}.
Proof: a constituent nodes equals projection deps r for some root r.
Every node in the projection is dominated by r, giving downward paths.
The catena BFS traverses edges bidirectionally, so any start node can
walk up to r (reversed dominance edges) then down to any target.