Conceptual dependency theory¶

Normalize paraphrastic natural-language inputs into canonical, language-independent event structures built from primitive acts, typed participants, dependencies, and inference-enabling causal expectations.

Version: v1 · 2026-08-30 · History
Domain-specific #: 1527
Origin domain: artificial intelligence
Subdomain: canonical event semantics
Aliases: Conceptual dependency, CD theory, Conceptual dependency representation

Core Idea¶

Conceptual dependency theory is Roger Schank's historical knowledge-representation framework for expressing the meaning of natural-language events in a canonical form intended to be independent of the particular words used. It decomposes actions into a small inventory of conceptual primitives, links them to actors, objects, directions, instruments, times, locations, and states, and uses dependency relations to support inference. Paraphrases should converge on one underlying representation when their asserted meaning is the same.^[1]

A parser identifies an event and selects a primitive act such as ATRANS for transfer of an abstract relation, PTRANS for physical movement, MTRANS for information transfer, PROPEL for applied force, or INGEST for taking something into an animate object. Typed slots and conceptual cases bind the participants and circumstances; causal and state-change dependencies encode consequences not always stated on the surface. Canonicalization reduces lexical variation, while attached inference rules enable the system to derive expectations used in understanding, question answering, memory, planning, and later script-based models.^[2]

The primitive inventory and notation changed across publications and implementations; it should not be presented as a current universal semantic ontology or an empirically complete decomposition of all events. Conceptual dependency differs from syntactic dependency grammar, which links words by grammatical relations, and from generic semantic networks, which do not require Schank's primitive-act canonicalization. Scripts were developed in the same research program but represent stereotyped event sequences at a higher level. Historical influence does not establish that the framework is the best modern natural-language architecture.^[3]

Structural Signature¶

Natural-language input. A sentence or discourse supplies a surface whose asserted event meaning is to be represented.
Canonicalization goal. Paraphrases with the same meaning are directed toward one underlying form.
Primitive acts. A small typed vocabulary such as ATRANS, PTRANS, MTRANS, and PROPEL decomposes events.
Participant roles. Actors, objects, recipients, sources, destinations, and instruments fill declared slots.
State and setting. Time, location, possession, physical state, and mental state constrain the event.
Dependency links. Causal, temporal, and conceptual relations connect acts and consequences.
Inference rules. The representation licenses unstated but framework-governed expectations.
Processing task. Parsing, memory, question answering, planning, or discourse understanding consumes the structure.

What It Is Not¶

Not syntactic dependency grammar. That analyzes grammatical head-dependent relations among surface words.
Not a generic semantic network. CD requires canonical primitive acts and inference-oriented role structure.
Not script theory. Scripts organize stereotyped event sequences and can be built above CD representations.
Not a universal ontology of reality. The primitive set is a designed historical theory with coverage and granularity choices.
Not mere synonym replacement. Canonicalization decomposes events and roles rather than substituting a preferred word.
Not a contemporary performance guarantee. Historical coherence and influence do not establish current benchmark superiority.

Scope of Application¶

The abstraction is literal wherever practitioners can identify the same constitutive roles, apply the same boundary tests, and obtain the same kind of output. The following habitats are uses of Conceptual dependency theory itself, not metaphors based only on resemblance.

Natural-language parsing. Mapping multiple verb forms and sentence structures into canonical event representations.
Paraphrase handling. Treating semantically equivalent inputs alike despite lexical differences.
Inference. Deriving possession, location, state-change, and causal expectations from represented acts.
Question answering. Answering from normalized event roles rather than matching surface words alone.
Narrative understanding. Connecting event representations with memory structures and later script theory.
History of AI. Studying a knowledge-intensive alternative to syntax-only and search-oriented processing.

Clarity¶

A clear account of Conceptual dependency theory must preserve the recognition invariant stated in the Core Idea rather than rely on the title alone. Identify the publication or implementation because primitive inventories and notation changed over time. Separate the input sentence, asserted meaning, chosen primitive, filled roles, and licensed inferences. Do not label a word-dependency parse or arbitrary semantic graph as conceptual dependency. Evaluate representational coverage, inference correctness, and parsing performance as separate questions. These declarations are not editorial extras: each changes what observations count, which transformations are licensed, and what conclusion can be drawn. A reader should be able to reconstruct the input, the operative rule, the output, and at least one defeater from the account without consulting an implementation or guessing an unstated convention.

Manages Complexity¶

Conceptual dependency theory manages complexity by replacing a diffuse field of observations or possible operations with a bounded role structure: natural-language input supplies a sentence or discourse supplies a surface whose asserted event meaning is to be represented.; canonicalization goal supplies paraphrases with the same meaning are directed toward one underlying form.; primitive acts supplies a small typed vocabulary such as ATRANS, PTRANS, MTRANS, and PROPEL decomposes events.; participant roles supplies actors, objects, recipients, sources, destinations, and instruments fill declared slots.; state and setting supplies time, location, possession, physical state, and mental state constrain the event.. The compression is useful because it localizes disagreement. One can ask whether the input was properly formed, whether a constitutive relation held, whether an alternative explanation defeats the inference, or whether the output was overinterpreted. The same compression can mislead when its discarded detail is exactly what the decision requires. A reference-grade use therefore reports both the invariant retained and the information intentionally lost.

Abstract Reasoning¶

Determine the asserted event meaning and resolve only the ambiguity justified by context.
Select the conceptual primitive or composition of primitives matching the event transition.
Bind actors, objects, directions, recipients, instruments, times, locations, and relevant states.
Normalize paraphrases by checking whether the resulting role and transition structure is identical.
Apply only the inference rules attached to the selected primitives and dependencies.
Compare predicted consequences with the discourse and revise an overcoarse or incorrect decomposition.
Retain uncertainty when the surface input underdetermines a unique conceptual dependency structure.
Test the candidate interpretation against the nearest named confusable rather than accepting a shared surface feature.
State the conclusion at the same scope as the source conditions, and retain uncertainty or nonuniqueness where the construct does not remove it.

Knowledge Transfer¶

The strict upward abstraction is Representation. Conceptual Dependency Theory instantiates Representation because it encodes event meaning in a structured surrogate whose primitives and relations make some inferences directly available while suppressing surface wording. Within canonical event semantics, the full mechanism transfers literally when the same roles and boundary tests recur. Beyond that domain, only the parent-level skeleton should travel. Reusing the label Conceptual dependency theory after removing its constitutive vocabulary would hide a change of mechanism behind an analogy. The honest transfer rule is therefore two-stage: recognize the domain-specific pattern first, then lift only the parent relation that remains invariant under a substrate change.

Examples¶

Canonical¶

The sentences 'John gave Mary a book' and 'Mary received a book from John' can be represented by the same ATRANS: possession of the book moves from John to Mary, with John as source, Mary as recipient, and the book as object. A merely physical carrying event would instead require a PTRANS component. The canonical form therefore turns paraphrase equivalence into an explicit role-and-transition test rather than a shared-keyword judgment.

Mapped back: input and conventions → constitutive role test → bounded output → explicit interpretation and defeater check.

Applied / In Practice¶

A historical question-answering system parses 'Alice told Bob the news' as an information-transfer act with Alice, Bob, and the conceptual object in typed roles. It can infer that Bob may now know the news, subject to the framework's assumptions. An analyst reports that as a licensed CD inference, not as proof that Bob believed, remembered, or acted on the message.

Mapped back: field observation or problem → candidate recognition → confusable and limit checks → appropriately scoped conclusion.

Structural Tensions¶

T1: Canonical form versus genuine ambiguity. Forcing one graph can erase unresolved word sense, scope, or reference. Diagnostic: What evidence selects this representation over the remaining alternatives?
T2: Small primitive inventory versus expressive coverage. Compression aids inference but can flatten domain-specific distinctions. Diagnostic: Which downstream inference changes if two acts share a primitive?
T3: Language independence versus designer vocabulary. The representation avoids surface words but uses theory-specific categories. Diagnostic: Would analysts from another linguistic framework choose the same decomposition?
T4: Inference power versus overgeneration. Default consequences can add facts not warranted in an exceptional context. Diagnostic: Which defeater blocks the inference, and can the system represent it?
T5: Historical influence versus present adequacy. Descendant ideas can survive after the original primitive set is abandoned. Diagnostic: Is a claim about CD itself or about a later knowledge-based descendant?
T6: Autonomy versus Representation. The parent models complex ideas; CD adds canonical primitive acts, typed dependencies, and inference rules. Diagnostic: Remove the primitive inventory and paraphrase-normalization test and see whether only generic representation remains.

Structural–Framed Character¶

The typed graph and primitive-act transition rules are structural within a specified version; sense resolution, primitive granularity, default inference, and evaluation criteria remain model-framed. The five framing criteria point in a consistent direction. Evaluative weight is limited to whether the defining conditions are met, not whether the outcome is desirable. Human practice matters to the extent that experts choose conventions, instruments, or reporting thresholds, but those choices do not make every verdict arbitrary. Institutional history explains the name and standard use; it does not replace the recognition rule. The operative vocabulary travels within the home field and closely adjacent subfields, while transfer farther away requires translation to the parent prime. Thus recognition remains disciplined even where interpretation is defeasible.

Structural Core vs. Domain Accent¶

What is skeletal. Conceptual Dependency Theory instantiates Representation because it encodes event meaning in a structured surrogate whose primitives and relations make some inferences directly available while suppressing surface wording. This is the part that can be expressed without the candidate's specialist nouns.

What is domain-bound. The irreducible accent is canonical event meaning, Schank's primitive acts, conceptual cases, dependency arrows, paraphrase normalization, attached state changes, and inference-oriented processing. Remove those elements and the result is no longer Conceptual dependency theory; it is only the parent relation or a loose analogy.

Why this does not clear the prime bar. The name does not recur with unchanged diagnostics across three independent domains. What transfers is already represented by prime:representation. The candidate remains autonomous because its in-domain recognition rule, failure modes, and consequences are stable, but its vocabulary and interventions do not float free of the home substrate.

Conceptual Dependency Theory instantiates Representation because it encodes event meaning in a structured surrogate whose primitives and relations make some inferences directly available while suppressing surface wording.

The prospective workspace queue contains one strict upward edge to prime:representation. No live DAG mutation is authorized.

Relationships to Other Abstractions¶

Current abstraction Conceptual dependency theory Domain-specific

Parents (1) — more general patterns this builds on

Conceptual dependency theory is a kind of Representation Prime

Conceptual Dependency Theory instantiates Representation because it encodes event meaning in a structured surrogate whose primitives and relations make some inferences directly available while suppressing surface wording.

Hierarchy path (1) — routes to 1 parentless root

Conceptual dependency theory → Representation → Abstraction

Neighborhood in Abstraction Space¶

Conceptual dependency theory sits in a sparse region of the domain-specific corpus (91^st percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Unclustered & Miscellaneous (1565 abstractions)

Nearest neighbors

Language Production — 0.80
Semantic parsing — 0.78
Functional discourse grammar — 0.78
Logic form — 0.78
Functional Generative Description — 0.78

Computed from structural-signature embeddings · 2026-09-08

Not to Be Confused With¶

dependency grammar. Represents syntactic relations among words rather than canonical event semantics.
semantic network. A broad graph formalism without CD's required primitive-act decomposition.
script. A stereotyped temporally ordered event package developed above individual CD structures.
case grammar. Organizes semantic or syntactic cases around predicates but uses a different theoretical apparatus.
frame semantics. Represents lexical frames and roles rather than Schank's fixed primitive-transition scheme.
conceptual graph. A distinct formal representation tradition despite the similar surface name.

References¶

[1] Schank, R. C., and Tesler, L. (1969). 'A Conceptual Dependency Parser for Natural Language.' COLING 1969, Preprint No. 2. ACL Anthology C69-0201. https://aclanthology.org/C69-0201/ registry ↩

[2] Schank, R. C. (1972). 'Conceptual Dependency: A Theory of Natural Language Understanding.' Cognitive Psychology 3(4), 552–631. https://doi.org/10.1016/0010-0285(72)90022-9 registry ↩

[3] Lytinen, S. L. (1992). 'Conceptual Dependency and Its Descendants.' Computers & Mathematics with Applications 23(2–5), 51–73. https://doi.org/10.1016/0898-1221(92)90136-6 registry ↩