Co-occurrence Analysis — Word Co-occurrence
Also known as: word co-occurrence, co-occurrence network, Kelime Eş-Oluşum Analizi
Co-occurrence analysis is a text-mining technique that statistically counts the word pairs that appear together within a window or a sentence and uses their frequencies to reveal semantic maps and thematic structure. It rests on the distributional principle articulated by J.R. Firth in 1957 — that a word is characterised by the company it keeps.
Key highlights
- Simple and interpretable — counts of word pairs are easy to understand and visualise as a network.
- Surfaces themes and concept clusters from unstructured text without requiring labelled data.
- Grounded in the well-established distributional principle that related words share contexts.
Intuition
This section is available to Pro members. Upgrade to Pro
How it works
This section is available to Pro members. Upgrade to Pro
When to use it
Use co-occurrence analysis when you have text data and want to explore relationships and thematic structure between terms rather than test a hypothesis. It suits exploratory and relational goals on a corpus of at least roughly thirty documents. It assumes you have removed stop-words, chosen a context window size, and preprocessed the text; without text data the method cannot run.
Strengths & limitations
- Simple and interpretable — counts of word pairs are easy to understand and visualise as a network.
- Surfaces themes and concept clusters from unstructured text without requiring labelled data.
- Grounded in the well-established distributional principle that related words share contexts.
- Results depend heavily on the chosen context window size and on the quality of stop-word removal.
- Raw frequency counts can be dominated by common terms unless preprocessing is careful.
- Captures association, not causation or directed semantic relations between the paired words.
Common pitfalls
This section is available to Pro members. Upgrade to Pro
Applications
This section is available to Pro members. Upgrade to Pro
Frequently asked
How should I choose the context window size?
There is no universal answer — it is a deliberate parameter. A wider window (e.g. a whole sentence or many tokens) captures looser topical association, while a narrower window captures tighter, more syntactic pairings. Choose it to match the kind of relationship you want to surface, and report the value you used.
Why is stop-word removal so important here?
Function words like 'the' and 'and' co-occur with almost everything, so without removing them the co-occurrence matrix is dominated by uninformative pairs that hide the real associations between content words.
Does high co-occurrence mean two words are truly related?
Not necessarily. Strong co-occurrence is evidence of association, but it can also reflect that both words are simply very frequent. It captures association rather than a directed or causal semantic relation, so interpret clusters with care.
How much text do I need?
The method expects a corpus of at least roughly thirty documents. With fewer, the co-occurrence counts become unstable and noisy, and the resulting map is unreliable.
Sources
- 1.Firth, J.R. (1957). A Synopsis of Linguistic Theory. Studies in Linguistic Analysis. Oxford: Blackwell.
- 2.Turney, P.D. & Pantel, P. (2010). From Frequency to Meaning: Vector Space Models of Semantics. Journal of Artificial Intelligence Research, 37, 141-188.
You have read it. What now?
Cite this page
ScholarGate. (2026, June 1). Co-occurrence Analysis. ScholarGate. https://scholargate.app/text-mining/co-occurrence-analysis