Death of the Novel(ty): Beyond N-Gram Novelty as a Metric for Textual Creativity

The paper operationalizes textual creativity as a dual construct combining novelty and appropriateness (sensicality plus pragmaticality), then empirically tests whether n-gram novelty alone suffices as a proxy. It resides in the 'Critique and Comparative Analysis of Creativity Metrics' leaf alongside three sibling papers that similarly scrutinize existing evaluation approaches. This leaf sits within the broader 'Creativity Evaluation Methodologies and Metrics' branch, which contains three leaves totaling twelve papers. The critique-focused direction is moderately populated, suggesting active debate over how to measure creative language without reducing it to surface-level statistics.

The taxonomy reveals neighboring leaves dedicated to 'Novel Metric Development' (four papers proposing new scoring systems) and 'Domain-Specific Creativity Evaluation' (four papers examining poetry, humor, and game content). The original paper bridges these areas: it critiques n-gram novelty (aligning with its own leaf's scope) while proposing a richer operationalization (touching on novel metric territory). Upstream, the 'Theoretical Foundations' branch (eight papers) establishes cognitive and philosophical grounding for creativity, which the paper invokes when defining appropriateness. Downstream, the 'Human-AI Creativity Comparison' branch (five papers) conducts empirical studies similar to the paper's LLM evaluation component, though those works focus on benchmarking rather than metric critique.

Among thirty candidates examined, none clearly refutes the three contributions. The first contribution—operationalizing creativity beyond n-gram novelty—examined ten candidates with zero refutable overlaps, suggesting that the dual-construct framing (novelty plus appropriateness via close reading) may be relatively novel within the search scope. The second contribution (expert writer annotation dataset) and third (close reading task for LLM evaluation) each examined ten candidates with zero refutations, indicating that the specific methodology of recruiting expert writers for fine-grained sensicality and pragmaticality judgments has limited direct precedent in the retrieved literature. These statistics reflect the bounded search, not exhaustive coverage.

Given the limited search scope of thirty semantically similar candidates, the paper appears to occupy a distinctive niche: it combines metric critique with a human-expert annotation study grounded in close reading, a methodology less common in the retrieved computational creativity literature. The taxonomy context shows that while metric critique is an active area, the specific operationalization and evaluation approach may differentiate this work from prior efforts. Broader literature beyond the top-thirty matches could reveal additional overlaps, particularly in creativity psychology or writing studies domains not fully captured by semantic search.