Inter-Rater Reliability of LLM and Rule-Based Annotation for Inferential Narrative Features: Three Studies on a Turkish Corpus

Datasets that ship automatically generated feature annotations invite a question that is rarely asked of them: would a human agree with those labels? This report answers that question for the Objective Projection corpus, a Turkish narrative dataset whose scenes carry a per-scene applied_rules field produced by a rule-based detector over six craft features — two prohibitions (explicit emotion labelling, simile) and four positive techniques (materialized metaphor, micro focus, temporal anchor, atmosphere contradiction). Three studies are reported. Study 1 (n = 120) scores the detector against blind labels produced by the scheme’s own author.

Study 2 (n = 100, a disjoint scene set) scores the detector plus Gemini 2.5 Flash and Grok against an independent non-expert human rater whose labels were locked before any machine ran. Study 2b re-runs the identical protocol with Claude Fable 5 (High) and ChatGPT 5.5. The central result concerns one rule. On materialized metaphor — the feature closest to the methodology’s theoretical core — the five machine labellers returned positive rates of 0, 1, 40, 72 and 78 out of 100 scenes, against a human count of 9.

Cohen’s κ was at or indistinguishable from chance for five of the six machine labellers, across both human references, on both scene sets: 0.004, 0.015, 0.000, 0.019, 0.027. Raw agreement, by contrast, ranged from 74.7% to 84.5%, an artefact of class imbalance rather than a sign of competence. We deliberately do not resolve the finding into a single story. Two readings survive the data: that the feature is genuinely inferential and beyond current automatic detection, or that the rule’s definition is not yet operational enough for any rater to apply consistently — including the human.

Distinguishing them requires a second independent human rater, which this report does not have and therefore does not claim.