Codebooks and validation annotations for README and paper-impact prediction

This dataset provides the codebooks and validation annotation data associated with the evaluation presented in the manuscript [under review]. The archive contains: README codebook (best-performing configuration): the final codebook for predicting GitHub repository popularity from README content, comprising 39 QA pairs and 128 answer codes, together with a validation set of 200 annotated GitHub repositories. Human-authored README codebook: an expert-drafted reference codebook of 22 questions (44 answer codes) for the README popularity task, used as a comparison baseline. Paper-impact codebook: the final codebook for predicting the expected impact of newly published papers from their titles and abstracts, comprising 32 QA pairs and 119 answer codes. Access to the files is currently restricted pending publication.