Measuring non-stationary semantic correlation in the dissemination effectiveness of mainstream ideology

In the social media era, mainstream ideology dissemination faces a key contradiction: information reaches many people, but whether its core meaning remains stable is hard to measure.Existing evaluations rely on behavioural metrics like forwards and likes, answering 'how many saw it' but not 'how many truly understood it'.During dissemination, texts are constantly rewritten, causing original semantics to drift or reverse.Current text similarity methods also lack adaptation to 'reference framework-specific' matching.To address this, this paper proposes the semantic association volatility indicator, transforming effectiveness evaluation into measuring semantic stability over time.Experiments on public corpora show our method achieves 74.2% accuracy in determining effectiveness levels -11.4 percentage points higher than static initial-moment methods.The semantic association volatility indicator also correlates at 0.83 with manual effectiveness labels, confirming non-stationary semantic association as a feasible proxy for dissemination effectiveness.