Self-Reported Height and Shoe Size Data from Forocoches

self_reports_fc.csv This repository contains a tabular dataset of 4,739 self-reported height and shoe-size entries collected from public discussion threads on Forocoches between 2010 and 2025. The dataset has been processed to remove any potentially identifiable or traceable information. All identifiers have been replaced with synthetic numeric IDs, and metadata that could enable reconstruction of original posts has been excluded. The dataset includes the following fields: thread — anonymized thread identifier (numeric ID) user — anonymized user identifier (numeric ID) height — reported height (cm) shoe_size — reported EU shoe size Height values are expressed in centimetres, and shoe sizes follow the European sizing system. The data reflect self-reported measurements as originally provided by users. No validation, plausibility filtering, or external correction has been applied. This dataset is intended for population-level statistical analysis of quantitative relationships derived from informal numeric self-disclosure. The anonymization procedure preserves structural relationships in the data while reducing the risk of re-identification. self_reports_fc_validation.csv This repository also includes a validation dataset containing 500 entries derived from the original corpus. The dataset has been processed following the same anonymization procedure as the main dataset. All identifiers have been preserved as in the original dataset to ensure traceability across datasets, and no information enabling the reconstruction of original posts is included. The dataset includes the following fields: thread — anonymized thread identifier (numeric ID) user — anonymized user identifier (numeric ID; may be missing, indicating that the user was not included in the dataset) height_human — height value manually extracted (cm) height_llm — height value extracted by the LLM (cm) shoe_human — shoe size manually extracted (EU size) shoe_llm — shoe size extracted by the LLM (EU size) height_detect_human — indicator (1/0) of whether a height value is present according to manual extraction height_detect_llm — indicator (1/0) of whether a height value is detected by the LLM shoe_detect_human — indicator (1/0) of whether a shoe size value is present according to manual extraction shoe_detect_llm — indicator (1/0) of whether a shoe size value is detected by the LLM height_error_abs — absolute difference between height_human and height_llm (cm) shoe_error_abs — absolute difference between shoe_human and shoe_llm height_category — classification outcome for height detection (TP: value present and correctly detected; TN: value not present and correctly not detected; FP: value not present but incorrectly detected; FN: value present but not detected) shoe_category — classification outcome for shoe size detection (TP, TN, FP, FN) This dataset is intended for evaluating the performance of automated extraction methods on informal numeric self-disclosure, by comparing manually extracted values with those obtained from the LLM. The manual validation followed a simple set of consistency rules: posts were considered valid if they contained at least one explicit numeric disclosure (height or shoe size); quoted content was not treated as a new disclosure, although any accompanying text was interpreted as a potential source of numeric information; and in cases where multiple values were reported, a single value was derived as the average of the explicitly stated values in the post.