DG Valency Bridge: [Osb19] #
Full derivation chain from DG valency theory to subcategorization and passive contrasts, grounded in the English Fragment lexicon.
Each verb's DG frame is DERIVED from its Fragment entry's complementType
field via complementToArgStr, not stipulated independently. The DG
analysis then verifies that trees built with these Fragment-derived words
satisfy the corresponding valency frames.
Derivation Chain #
Fragment VerbEntry.complementType ← lexical data (sleep=.none, kick=.np, give=.np_np)
↓ complementToArgStr
ArgStr frames (argStr_V0/VN/VNN) → DepTree.frames premises
↓
DepTree instances ← concrete parse trees
↓
satisfiesArgStr / checkVerbSubcat ← frame satisfaction + subcat verification
↓
passiveRule (LexRule) ← valency change derivation
↓
isCatena / isConstituent ← structural analysis (Ch 4)
↓
grammaticality contrasts ← predictions for the example sentences
sleep.complementType =.none → intransitive frame (V0).
devour.complementType =.np → transitive frame (VN).
give.complementType =.np_np → ditransitive frame (VNN).
kick.complementType =.np → transitive frame (VN, active).
The passive frame carried on passiveTree.frames is consistent with the
passive rule: the rule removes the obj slot from kick's transitive frame.
"John sleeps" — intransitive, no object.
Equations
Instances For
"John devours pizza" — transitive with object.
Equations
Instances For
"John gives Mary book" — ditransitive with two objects.
Equations
- One or more equations did not get rendered due to their size.
Instances For
"John kicked the ball" — active transitive (for passive derivation).
Equations
- One or more equations did not get rendered due to their size.
Instances For
"The ball was kicked" — short passive; the passive analysis derives
argStr_VPassive from kick's transitive frame (passive_frame_matches).
Equations
- One or more equations did not get rendered due to their size.
Instances For
"The ball was kicked by John" — long passive with agent by-phrase.
Equations
- One or more equations did not get rendered due to their size.
Instances For
"*John sleeps book" — intransitive with spurious object.
Equations
- One or more equations did not get rendered due to their size.
Instances For
"*John devours" — transitive missing required object.
Equations
Instances For
"*John gives Mary" — ditransitive missing direct object.
Equations
- One or more equations did not get rendered due to their size.
Instances For
"*The ball was kicked the ball" — passive with spurious object.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Intransitive tree satisfies intransitive frame (V0).
Transitive tree satisfies transitive frame (VN).
Ditransitive tree satisfies ditransitive frame (VNN).
Intransitive tree does NOT satisfy transitive frame (missing obj).
Transitive-minus-object tree does NOT satisfy transitive frame.
The passive rule applies to transitive "kicked".
The derived entry removes the obj slot and adds optional obl.
The passive rule output matches argStr_VPassive.
SVO tree: verb + all args {0, 1, 2} is a constituent (complete subtree).
SVO tree: verb + subject {0, 1} is a catena (connected via nsubj).
SVO tree: verb + subject {0, 1} is NOT a constituent (missing obj).
Ditransitive: verb + all args {0, 1, 2, 3} is a constituent.
Ditransitive: verb + just obj {1, 3} is a catena but not constituent.
Full valency derivation chain: from Fragment lexicon through DG theory to grammaticality predictions.
- Fragment kick.complementType =.np → transitive frame (argStr_VN)
- Active tree satisfies transitive frame (argStr_VN) ✓
- checkVerbSubcat validates the active tree ✓
- Passive rule applies and removes obj slot ✓
- Passive tree satisfies derived frame (argStr_VPassive) ✓
- checkVerbSubcat validates the passive tree ✓
- Passive + spurious obj correctly rejected ✗
DG subcategorization predictions capture the textbook contrasts:
| Grammatical | Ungrammatical | Frame |
|---|---|---|
| "John sleeps" | "*John sleeps book" | argStr_V0 |
| "John devours pizza" | "*John devours" | argStr_VN |
| "John gives Mary …" | "*John gives Mary" | argStr_VNN |
DG passive prediction captures the contrast "the ball was kicked" / "*the ball was kicked the ball".