ECHO-PM Dataset: Preprocessed GitHub Messages for Sentiment, Emotion, and Empathetic Communication Experiments

ECHO-PM Dataset (v1.0): Preprocessed GitHub Messages for NLP Experiments in Software Engineering Communication This record provides a dataset of 677 textual messages collected from public GitHub repositories and preprocessed with spaCy, intended to support Natural Language Processing (NLP) experiments in technical communication within Software Engineering contexts (e.g., exploratory analyses, message characterization, preparation for classification pipelines, and reproducibility of preprocessing steps). Included files base_dados_dissertacao_prep_spacy.csv (main dataset) A dataset datasheet (PDF) describing the dataset LICENSE.txt (CC BY 4.0) Scope of this release (v1.0) Total records: 677 Repositories: pallets/flask (343) and psf/requests (334) Message types: comment (556) and issue_body (121) Text representations: raw text + cleaned text + tokens + lemmas Rigor note (important): this CSV does not include labels (e.g., sentiment/emotion/tone). If labeled annotations or model outputs are released, they will be provided as separate files (e.g., labels_v1.csv) with a clearly documented labeling protocol and schema. Main CSV columns (high level) Metadata: repo, issue_number, kind, author, created_at, title, comment_id Text: message, clean_text, tokens, lemmas Measures/flags: tamanho, num_palavras, flag_long, flag_log, flag_code, flag_descartar(Note: in this release, the flags are present but not activated.) Intended use This dataset is designed to: enable reproducibility of a standardized preprocessing pipeline, and serve as a base for constructing labeled subsets and downstream modeling/evaluation pipelines. Ethics & privacy All content originates from public GitHub sources. The author field contains public GitHub handles; for stricter privacy requirements, consider releasing an anonymized variant (hash/salt) in a future version. License Released under Creative Commons Attribution 4.0 International (CC BY 4.0). Citation DOI: 10.5281/zenodo.17991751 Suggested citation (plain text):Barros, F. A. S. (2025). ECHO-PM Dataset (v1.0): Preprocessed GitHub Messages for NLP Experiments in Software Engineering Communication. Zenodo. https://doi.org/10.5281/zenodo.17991751