Do all verbs allow the dative alternation?

No. Acceptability depends on the verb, sense, construction and language variety. A rule learned from a few examples should not be universalised.

Do alternations always preserve meaning exactly?

No. They can preserve central participant relationships while changing focus or adding and removing explicit event information. The precise relationship must be checked.

AI Glossary / Natural language and speech / Intermediate

Verb alternations

Definition

Definition: Verb alternations are related ways in which a verb's participants can be expressed in sentence structure. For example, “Nora sent the file to Omar” and “Nora sent Omar the file” place the transferred item and recipient in different grammatical positions while describing a closely related transfer event.

What it means

What it means: The participants in an event are not tied to one fixed word order or grammatical slot. A thing affected by an event can appear as an object in one construction and a subject in another. This makes verb alternations important for systems trying to identify who did what to whom.
The term concerns argument structure and its realization, not merely changing send to sent. Inflection changes a word's form; an alternation changes how participants are expressed. The distinction also does not mean that every alternation preserves exactly the same meaning, emphasis or assumptions.

How it works

How it works: Linguists and computational resources describe the syntactic frames that a verb can occur in and relate them to semantic roles. A dative alternation can express a recipient in a to-phrase or as an object preceding the transferred item. A causative/inchoative alternation can contrast “Nora opened the gate” with “The gate opened,” changing whether a cause is explicitly expressed.
Participation is verb- and construction-dependent. A pattern that works for give or send cannot simply be applied to every verb with a vaguely similar meaning. Resources such as VerbNet organize verbs into classes with frames, roles and semantic descriptions. Their entries are useful evidence, but applying them still requires the intended verb sense and the sentence context.

Three practical examples

Three practical examples:
1. A message extractor should identify Omar as recipient in both “Nora sent Omar the file” and “Nora sent the file to Omar.” Treating the first noun after every verb as the transferred item would fail on one construction.
2. A maintenance log says “The latch broke.” A system can identify the affected object but should not invent a human agent merely because another sentence could say “A technician broke the latch.”
3. A writing assistant considers an alternative phrasing. It checks whether the specific verb permits the proposed construction and whether the change preserves the intended emphasis. Mechanical slot swapping can produce awkward or unacceptable sentences even when the individual words are correct.

Think of it this way

Think of it: The same delivery can be entered into forms that put recipient and item fields in different orders. A careful reader follows field roles rather than assuming that the first blank always means the same thing. Language is less rigid than a form, however: changing a construction can also change what is asserted, highlighted or left unspecified.

Worked example

Worked example: Use two invented records: A, “Nora sent the file to Omar”; B, “Nora sent Omar the file.” For this exercise, the target fields are sender, item and recipient. Both records should yield sender=Nora, item=the file and recipient=Omar.
A naïve rule selects the first noun phrase after sent as the item. It is correct for A and wrong for B, giving item-field accuracy of 1/2 across the two records. If the rule also assumes that a recipient must follow to, it finds the recipient in A but misses B. Recognizing the two frames permits the same role mapping for both intended readings.
This does not license swapping arbitrary verbs into the pattern. Nor does successful extraction prove that the transfer actually happened: quoted, negated or hypothetical versions require further interpretation. The example isolates argument realization rather than the full problem of factual event extraction.

Illustrative case study

Illustrative case study: Fictional Summit Service analyses equipment logs. Its initial rule equates grammatical subjects with responsible people. It consequently treats “The valve opened” as if the valve were an intentional actor and misses the distinction from “The operator opened the valve.”
The team separates grammatical relations from event roles and represents an unexpressed cause as unknown rather than filling in a guessed person. Reviewers test active, passive and inchoative descriptions alongside negated examples. This fictional case illustrates a representation improvement. It does not claim that semantic-role software can determine responsibility or recover information absent from the record.

Common misconceptions and mistakes

Common misconceptions and mistakes: Alternations are not unrestricted paraphrase rules. Similar verbs can differ in which frames they allow, and language varieties can differ in acceptability. A subject is not always an agent. An omitted participant is not necessarily nonexistent, but it is also not automatically recoverable. Related constructions may differ in emphasis, affectedness or event structure. Finally, two sentences with the same verb are not necessarily an alternation pair if they use unrelated senses or describe unrelated participant relationships.

Business use

Business use: Event extraction, knowledge-base population, search and translation benefit from recognizing multiple grammatical realizations of related events. Build evaluation sets containing the main constructions found in the business domain, rather than testing only one canonical word order. Preserve uncertainty when a participant is not expressed. Record the role schema used by the application because labels such as theme, patient, goal and recipient differ across resources. Accurate role mapping supports analysis, but business conclusions still require context and evidence.

Check your understanding

Check your understanding: From “The door opened,” a system records “A person opened the door.” Is that inference justified by knowing the causative/inchoative alternation?

Answer: No. The original describes the door's change of state without naming a person or asserting that a person caused it. Knowledge of a related construction does not supply missing facts.

Related terms

Dative alternation
Semantic role
argument structure
diathesis alternation
subject
object
causative
inchoative
VerbNet

Learn next

compare two sentences that use the same verb in different frames. Label grammatical positions and semantic roles separately. Then find a superficially similar verb for which the same transformation is unsuitable, and explain why lexical and constructional restrictions matter.

Sources

Sources: Jurafsky and Martin, Speech and Language Processing, Chapter 22, diathesis alternations and semantic-role distinctions: https://web.stanford.edu/~jurafsky/slp3/22.pdf . University of Colorado, VerbNet project description, verb classes, syntactic frames and semantic representations: https://verbs.colorado.edu/verbnet/ . The Nora/Omar extraction example and Summit Service case are original illustrations; no universal transformation rule or actual causation is inferred from them.

Previous: Verb Next: Verb phrase