AI Glossary / Natural language and speech / Intermediate
Allen relations
Definition
Allen relations are thirteen basic ways that two time intervals can relate through their beginnings and endings. They distinguish patterns such as before, meets, overlaps, during, and equals. Named for James Allen, they support qualitative temporal reasoning: a system can reason about how events are ordered or contained even when exact clock times are unavailable.
What it means
The standard framework assumes intervals with positive duration on a common linear timeline. For two intervals with definite endpoints, exactly one basic relation holds. Six relations have distinct converses, and equals is its own converse, giving thirteen in total.
The everyday word “overlap” is broader than the specific Allen relation overlaps. In ordinary scheduling, one event contained inside another overlaps it in time. In Allen's basic classification, strict containment is during or contains instead. Precise terminology prevents an application from mixing these meanings.
How it works
Compare the intervals' start and end points. A is before B if A ends strictly before B begins. A meets B if A ends exactly when B begins. A overlaps B when A starts first, B starts before A ends, and A ends before B ends.
A starts B if they begin together but A ends earlier. A finishes B if A begins later but they end together. A is during B if both its endpoints lie strictly inside B. Equals requires matching starts and matching ends. Reversing the interval order gives after, met by, overlapped by, started by, finished by, and contains.
If the evidence is incomplete, a relation can be represented as a set of possible basic relations. Composition combines constraints through an intermediate interval, sometimes narrowing the possibilities without determining a single result. Such possibilities are logical alternatives, not probabilities unless a separate probabilistic model is supplied.
Three practical examples
1. Meeting analysis: An assistant can distinguish a session that ends exactly as another begins from sessions separated by a gap, which matters when planning transition time.
2. Maintenance records: A report may establish that a fault occurred during a maintenance window without giving exact timestamps. A temporal representation can preserve that containment claim.
3. Document timelines: An information-extraction system can connect events described as preceding, overlapping, or continuing beyond one another, while retaining uncertainty where the text gives only partial ordering.
Think of it this way
Think of two colored bars on a timeline. Their relative endpoints tell you whether they touch, cross, nest, or coincide. This captures the basic geometry of the relations. The analogy does not establish causation, resource availability, or feasibility: two bars can fit neatly while the real tasks still require unavailable staff or equipment.
Worked example
Use an invented schedule on one date and in one time zone. A runs from 09:00 to 09:30. B runs from 09:30 to 10:00. C runs from 10:00 to 10:20. A meets B, and B meets C. A is before C, because A ends at 09:30 and C begins at 10:00; the thirty-minute interval B lies between them.
Now let D run from 09:15 to 09:45. A overlaps D: 09:00 is before 09:15, which is before 09:30, which is before 09:45. The shared time lasts fifteen minutes, but that numeric duration is extra information beyond the qualitative relation.
Let E run from 09:05 to 09:20. E is during A, because both its endpoints are strictly inside A. If E instead began at 09:00 while still ending at 09:20, E would start A rather than be during A under the strict basic definitions.
The calculations use exact invented endpoints. If a log rounds timestamps to the nearest minute, apparently equal boundaries may conceal a small gap or overlap. The representation must reflect the precision of the evidence rather than treating rounded equality as unquestionable physical simultaneity.
Illustrative case study
Fictional case: Cedar Operations combines maintenance notes with a calendar. Its first rule labels every pair sharing any time as the basic relation overlaps, including cases of complete containment.
The team replaces that rule with the full endpoint classification and keeps a separate broad “shares time” flag for its scheduling interface. Reviewers also distinguish reported ordering from inferred ordering. The case illustrates a terminology and modeling correction; it reports no deployed results or measured improvement.
Common misconceptions and mistakes
Do not infer that A caused B merely because A occurred before B. Temporal relations describe ordering, not causal mechanisms. Also avoid assuming that before specifies how long the gap is.
Do not apply the thirteen-way partition unchanged to zero-duration points. The standard interval assumptions matter, and point-event handling needs an explicit extension or convention. Finally, uncertain relation sets should not be silently collapsed to one guess, and local consistency checks should not be described as a universal solution to every interval-constraint problem.
Business use
Allen relations help make event timelines and scheduling logic more precise. They are particularly useful when evidence comes from text that supplies relationships but incomplete dates. A system can retain “during the review” without inventing a start time.
For operational use, document timezone handling, endpoint precision, interval validity, and the interpretation of touching boundaries. Keep temporal reasoning separate from resource scheduling and permission to act. A relation saying that two tasks meet does not establish that a person can travel between their locations instantly.
People also ask
Why are there thirteen relations?
Six directed patterns have distinct converses, giving twelve, and equality adds one self-converse pattern.
Does meets mean the intervals share a positive duration?
No. It means one ends where the other begins; their interiors do not overlap.
Can a relation remain uncertain?
Yes. A set of possible basic relations can represent incomplete information without selecting an unsupported exact relationship.
Check your understanding
A ends at 11:00 and B begins at 11:00, with both intervals having positive duration. Is A before B in the strict basic classification?
Answer: No. A meets B. Before requires A's end to be strictly earlier than B's start.
Related terms
- Interval algebra
- Event extraction
- Constraint propagation
- Composition
- temporal reasoning
- converse relation
- temporal ordering
- timestamp
Learn next
Draw endpoint diagrams for all thirteen relations, then explore relation composition. Start with meets followed by meets, and distinguish what follows logically from what would require additional clock-time information.
More in Natural language and speech
- Affix grammar over a finite lattice - Definition: An affix grammar over a finite lattice, usually abbreviated AGFL, is a grammar formalism that enriches context-free rules with features whose...
- Affricate sound
- Aktionsart - Aktionsart refers to the temporal character expressed by a verb or larger predicate: whether it describes a state, an ongoing activity, a process with a...
- Alveolar sound - An alveolar sound is a speech sound articulated with the tongue at or near the alveolar ridge, the raised area just behind the upper front teeth.
- Anaphor - An anaphor is an expression interpreted as referring back to something already introduced in a discourse. In “A parcel arrived.
- Anaphora - Anaphora is the use of a later expression to refer back to something introduced earlier in a discourse. In “A courier arrived.
Sources
Source: Thomas Alspaugh's author-maintained university account of Allen's Interval Algebra, read for definitions, converses, uncertainty sets, and composition: https://ics.uci.edu/~alspaugh/cls/shr/allen.html . Direct retrieval of Allen's original paper and the supplied textbook full-book URL was unavailable. Schedule examples are original; the entry presents the standard positive-duration interval framework.
Previous: All-to-all (parallel pattern) Next: Allocation concealment