Semantic Network Analysis
Also known as: Text network analysis, Concept co-occurrence network analysis, Centering resonance analysis, Anlamsal Ağ Analizi
Semantic network analysis represents the meaning of a text or corpus as a network of concepts connected by their co-occurrence or grammatical proximity, then uses network-analytic measures to reveal which ideas are central, how concepts cluster, and how shared meaning is structured. In communication research it is the standard way to map the conceptual architecture of media coverage, organizational discourse, and public conversation at scale.
Key highlights
- Scales interpretation to large corpora, surfacing the conceptual skeleton no analyst could read by hand.
- Imports the full toolkit of network measures — centrality, clustering, cohesion — into the analysis of meaning.
- Produces intuitive visual maps that communicate discourse structure to non-specialists.
- Supports rigorous comparison of conceptual structures across sources, time, languages, or communities.
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 semantic network analysis when you want to map the conceptual structure of a body of text — the central ideas, their clustering, and the bridges between themes — especially across a corpus too large to read closely. It suits questions about shared meaning, discourse organization, and how conceptual structures differ across outlets, communities, or periods, and it underpins network agenda-setting and computational framing work. It assumes that co-occurrence is a reasonable proxy for semantic association and that the extracted concepts capture the meaningful vocabulary. It is less appropriate when meaning hinges on sentence-level argument, negation, or narrative sequence that co-occurrence flattens, or when the corpus is small enough that close interpretive reading would be richer.
Strengths & limitations
- Scales interpretation to large corpora, surfacing the conceptual skeleton no analyst could read by hand.
- Imports the full toolkit of network measures — centrality, clustering, cohesion — into the analysis of meaning.
- Produces intuitive visual maps that communicate discourse structure to non-specialists.
- Supports rigorous comparison of conceptual structures across sources, time, languages, or communities.
- Co-occurrence ignores negation, irony, and argument structure, so 'linked' concepts may not be meaningfully associated.
- Results are sensitive to preprocessing choices — concept extraction, window size, thresholds — that can change the network substantially.
- Large networks become visually and analytically cluttered, requiring thresholds that discard information.
- It describes structure, not the propositional content of relations; an edge says two concepts co-occur, not what is asserted about them.
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 is semantic network analysis different from topic modeling?
Topic modeling (e.g., LDA) discovers latent topics as probability distributions over words and assigns documents to topics. Semantic network analysis instead represents concepts as nodes and their co-occurrence as edges, then uses network measures to study structure — which concepts are central, how they cluster, what bridges themes. Topic models answer 'what themes exist,' while semantic networks answer 'how are concepts related and organized,' and the two are often used together.
What is centering resonance analysis?
Centering resonance analysis (CRA), developed by Corman and colleagues, is a principled way to build semantic networks. Grounded in centering theory from linguistics, it selects the noun phrases that are discursively central within sentences and links them, then uses network influence measures to index each word's importance. CRA gives a theory-based alternative to ad hoc keyword selection and produces networks whose central nodes correspond to the text's organizing concepts.
Does the choice of co-occurrence window matter?
Yes, substantially. Linking concepts only within the same sentence yields sparser, more precise networks; using a wide sliding window or whole-document co-occurrence yields denser networks that may connect unrelated ideas. Because the window directly shapes which edges exist, best practice is to justify the window from the research question and to report whether key findings are robust to reasonable alternative window sizes and thresholds.
Sources
- 1.Corman, S. R., Kuhn, T., McPhee, R. D., & Dooley, K. J. (2002). Studying complex discursive systems: Centering resonance analysis of communication. Human Communication Research, 28(2), 157–206.
- 2.Doerfel, M. L., & Barnett, G. A. (1999). A semantic network analysis of the International Communication Association. Human Communication Research, 25(4), 589–603.
You have read it. What now?
Cite this page
ScholarGate. (2026, June 22). Semantic Network Analysis. ScholarGate. https://scholargate.app/communication/semantic-network-analysis