Abstract
National curricula regulate educational expectations and assessment through their statement genres and the linkage between expectations and criteria. This article compares Kazakhstan and Finland through cross-lingual computational document analysis of formal curriculum and assessment-regulation texts; it does not examine classroom assessment or implementation. Atomic normative statements were embedded and compared using directional mean-best similarity, threshold coverage, one-to-one matching, and size-matched bootstrapping. Both systems show strong internal textual coherence through different architectures: Kazakhstan exhibits close outcomes-to-criteria coupling, whereas Finland links a broader objective layer to criteria. Cross-system convergence is higher for expectations/objectives than for criteria, indicating that evaluative phrasing is more system-specific than curricular intentions. Finland shows higher semantic alignment with literacy, civic and social learning, sustainability, and inclusion frames, whereas Kazakhstan shows higher semantic alignment with digital-competence language in the assessment-related layer. The findings show that global curriculum discourses travel more readily at the level of stated aims than assessment rules, where evaluative frameworks remain less portable.