The words of a query carry part of the meaning. The situation carries the rest.

Pragmatics is the branch of linguistics that studies how the context of an utterance contributes to its meaning. The same words mean different things in different situations, and the difference is orderly enough to be described. A query is an utterance, so everything pragmatics established applies to it.

“Pragmatics is the study of linguistic acts and the contexts in which they are performed”

Reference: Stanford Encyclopedia of Philosophy, Pragmatics. https://plato.stanford.edu/entries/pragmatics/

Semantics describes what the words of a sentence mean by convention. Pragmatics describes what a speaker means by uttering them here, now, to this listener. The sentence do you have any spare change has one conventional meaning and several speaker meanings, and the situation picks one.

“more is involved in what one communicates than what one literally says”

Reference: Stanford Encyclopedia of Philosophy, Pragmatics. https://plato.stanford.edu/entries/pragmatics/

Speaker meaning is what a person intends to communicate, as distinct from the standing meaning of the words. Paul Grice introduced the distinction in 1957 and built the modern field on it. A search system receives the words and has to infer the rest.

Deixis is the use of expressions whose reference depends on the situation of utterance: here, now, this, you, near me. A query full of deictic words is readable only with a situation attached, which is why best restaurant near me means one thing in Frankfurt at eight in the evening and another in Hamburg at noon.

“is the use of words or phrases to refer to a particular time”

Reference: Wikipedia, Deixis. https://en.wikipedia.org/wiki/Deixis

Grice described conversation as a cooperative undertaking in which speakers say as much as is needed, no more, and keep what they say relevant, true and clear. Listeners assume this and infer accordingly. A searcher types three words and expects the system to behave like a cooperative partner who fills in the rest.

Common ground is what the parties to an exchange treat as shared and settled. Between two colleagues it is thick; between a searcher and a search system it is thin, and the searcher usually assumes it is thicker than it is. Most disappointing results come from that gap.

Four, and each has its own page here. Presupposition in search queries is what the query takes for granted. Conversational implicature: what queries and pages mean without saying it is what it suggests without saying. Speech act theory: what a query does are what it does, beyond describing. Relevance theory and what relevance means in search is the reader's balance of effect against effort. Taken together they account for the part of a query its words leave out.

Charles W. Morris divided the study of signs in 1938 into syntactics, semantics and pragmatics, and pragmatics was the part concerned with the relation between signs and their users. Linguistics adopted the division, and the field grew from Austin, Grice and Searle into the discipline it is now.

Speech acts have. Dialogue act annotation is published as an international standard, ISO 24617-2, so a machine can record what an utterance does in a way another machine can read. Presupposition, implicature and relevance are described in the literature and carried into systems by hand, one design decision at a time. The gap between those two states is where most answer quality is decided.

The system receives the words and has to reconstruct the situation. It has fragments to work with: place, time, device, the queries that came before, the language used. Everything else it assumes. Conversational systems make the problem visible by taking a sequence of turns, where each turn changes what the next one means.

“Conversational information seeking”

Study: Zamani, Trippas, Dalton and Radlinski, Conversational Information Seeking, 2022. https://arxiv.org/abs/2201.08808

Its own names, and they are older than the search industry. Patrick Wilson published situational relevance in 1973: the worth of an item measured against the question a person actually carries. Context-aware information retrieval adjusts ranking by the circumstances of the search. Extralinguistic context, user location and temporal anchor are the system-side names for the situation of utterance, for place deixis and for time deixis. One subject, two vocabularies, and the two rarely cite each other.

“Situationally relevant items of information are those that answer, or logically help to answer, questions of concern”

Study: Patrick Wilson, Situational Relevance, Information Storage and Retrieval 9(8), 1973. https://eric.ed.gov/?id=EJ084928

It names the assumption behind the query in plain words, gives the answer for the case where the assumption holds, and gives the answer for the case where it fails. It says who it is written for and when it applies. A page built that way holds up when the query arrives with a different situation attached, and it survives being summarised, because the conditions travel with the answer.

State the conditions. A sentence that begins if you are looking at this before signing gives the reader a test for whether the answer is theirs. It also gives an answer system something it can carry over: the condition belongs to the sentence, so it travels with the quote.