AI Glossary / Graphs and symbolic AI / Intermediate
Time and tense
Definition
Time and tense concerns how language describes when events or states hold, and how an AI system represents that temporal meaning. Time is the ordering or location of situations; tense is a grammatical way of relating them to a temporal viewpoint.
What it means
A sentence can describe a past event, an ongoing state, a future plan or a repeated activity. Understanding its nouns and verbs is not enough if the system assigns the situation to the wrong period.
Compare “the package arrived,” “the package is arriving” and “the package will arrive.” The same package and broad event type appear, but the claims differ. A useful representation keeps that temporal information connected to the event instead of reducing all three sentences to an undated arrival fact.
Tense does not provide an exact timestamp by itself. Words such as yesterday, before and next week, along with the surrounding conversation, often supply essential information. Grammatical systems also differ across languages, so an English tense label is not a universal temporal representation.
How it works
A language system identifies events and states, recognizes tense and temporal expressions, and determines the relevant reference point. It then represents relations such as before, after or overlapping, sometimes with explicit dates or intervals.
A helpful distinction separates event time, speech or writing time, and reference time. In “By Friday, the team had finished,” the finishing precedes a Friday reference point, which is itself viewed from a later reporting perspective. The exact completion instant may remain unknown.
Aspect adds information about how a situation is viewed, such as ongoing or completed, and interacts with tense. Modality matters too: a planned or predicted event is not an observed occurrence. A system should preserve unresolved alternatives when the text does not identify a unique date, rather than filling the gap with an invented timestamp.
Three practical examples
- A support assistant distinguishes an outage that ended yesterday from one that is still occurring when summarizing a conversation.
- A meeting-note tool interprets “tomorrow” using the original meeting date, rather than the date when the notes are later imported.
- A translation system preserves the ordering between an earlier delivery and a later inspection even when the languages use different grammatical forms.
Think of it this way
Think of placing labeled cards on a calendar, with arrows for relationships that are known even when exact dates are missing. A card saying “inspection after delivery” can preserve useful information without claiming a precise inspection time.
The analogy is limited because natural language also conveys possibility, intention, habit and viewpoint. A calendar position alone cannot capture the difference between an event that happened and one that someone merely expected to happen.
Worked example
Anchoring a relative expression
Consider an invented message written on 12 October: “We completed the test yesterday. By Friday, we had already packed the samples.” Assume the local date of writing is known and yesterday uses that local calendar. The test completion is therefore located on 11 October, though its time of day is unspecified.
The second sentence gives a different kind of information. Packing occurred before the Friday reference point. Unless the conversation identifies which Friday, the system cannot safely assign its calendar date. It also cannot assume that packing occurred after the test merely because the packing sentence appears second.
A careful record therefore stores the dated test event, a packing-before-Friday relation, and an unresolved Friday reference. Asking for clarification or retaining the uncertainty is more faithful than inventing a fully ordered story. This example deliberately distinguishes text order, event order and date resolution.
Illustrative case study
Fictional case: Alder Service builds a summarizer for equipment-support messages. A note says that a customer “was experiencing interruptions,” followed by a later message saying service “is stable now.” The first version reports an active interruption because it extracts the problem phrase without its temporal context.
The team adds evaluations that distinguish past problems, ongoing conditions and proposed future checks. It also preserves the timestamps of original messages and links each status claim to its source. When imported notes contain quoted older messages, the system treats the quotation's viewpoint separately from the import time.
The revised workflow gives reviewers the event and its temporal context together. This fictional case illustrates an evaluation need; it does not claim that a particular model or grammar resolves every ambiguity.
Common misconceptions and mistakes
- Past tense identifies an exact date. It usually provides a relative temporal relationship; other evidence supplies dates.
- The order of sentences is always the order of events. Narratives can look backward, quote earlier statements or describe parallel events.
- Future wording proves that an event will occur. It may express an intention, forecast, obligation or conditional possibility.
Business use
Temporal interpretation matters when AI updates records, extracts commitments or reports operational status. A system that correctly identifies an event but misplaces it in time can still produce a misleading summary.
Evaluation should include relative dates, quotations, changes of viewpoint, repeated events and statements whose dates remain unknown. Keep the original wording and document date alongside normalized values. For multilingual work, check meaning across languages rather than assuming that tense categories can be transferred one for one.
People also ask
Is tense the same as aspect?
No. Tense concerns temporal location relative to a viewpoint; aspect concerns how the situation is presented internally, although grammatical forms can combine them.
Can a present-tense form describe the future?
Yes. For example, “the train leaves tomorrow” has a future interpretation supplied by the whole expression.
Must every event receive a precise timestamp?
No. An interval, an ordering relation or an explicitly unresolved time can represent the available evidence more accurately.
Check your understanding
A message written on 8 May says, “The review finished yesterday.” It is imported on 12 May. Which date should anchor yesterday, assuming the original local date is reliable?
Answer: 8 May, giving 7 May as the review date. Importing the message does not change its original temporal viewpoint.
Related terms
- Event extraction
- Semantic analysis (computational)
- Aspect
- Natural language understanding
- Temporal reasoning
- semantic analysis
- reference time
- temporal interval
Learn next
Read temporal reasoning and event extraction. Take a short message containing yesterday, already or tomorrow, and mark which temporal facts are explicit, which follow from context, and which remain unknown.
More in Graphs and symbolic AI
- Thrashing - Thrashing is a pattern in which a system repeatedly discards and restores information, spending substantial effort on recovery instead of useful progress.
- Threshold function - A threshold function converts a numerical input into one of two outputs by comparing it with a cutoff.
- Tiling - In the neural-network context used here, tiling is a constructive learning method that grows a feedforward network by adding units and layers.
- Time line - A time line, usually written timeline, is an ordered representation of events, states or activities in time.
- Time to answer - Time to answer is the elapsed time a user needs to obtain a satisfactory answer to an information need.
- Tit-for-tat - Tit-for-tat is a strategy for a repeated two-player interaction: cooperate on the first round, then copy the other player's action from the preceding round.
Sources
Anna Kibort, Surrey Morphology Group, grammatical features inventory on tense: https://www.smg.surrey.ac.uk/features/morphosemantic/tense/
Nelma Moreira, Semantic Analysis of Time and Tense in Natural Language: an implementation: https://www.dcc.fc.up.pt/~nam/resources/publica/time.pdf
Relevant discussions of temporal viewpoints, linguistic forms and interval-based semantic representation were read. The business cases and date exercises are original examples, not reported system results.
Previous: Tiling Next: Time-aware long short-term memory