Replication_bundle

This deposit contains the replication materials for the manuscript "Virtuous and Vicious Polarization: Moral Segmentation and Democratic Backsliding in the Discourses of Correa and Uribe", currently under review. The bundle includes the full corpus of 79 presidential speeches (320,848 words) by Rafael Correa (Ecuador, 2007-2017) and Álvaro Uribe (Colombia, 2002-2010), the inductively constructed dictionary of moral-denunciation vocabulary (157 lemmas distributed across 8 semantic fields), the Python scripts that reproduce all quantitative findings reported in Tables 1 to 3, the four formal robustness tests reported in Appendix D (leave-one-speech-out, leave-one-field-out, leave-one-lemma-out, and cross-validation with a strict-core dictionary that excludes ten contextually sensitive lemmas), and the full methodological log documenting the six-step construction of the dictionary (Appendix B). Contents are organised in four directories. (i) "code" contains four executable Python files: 01_extract_text.py for PDF/DOCX extraction, MD5 deduplication, and chronological reassignment; 02_apply_dictionary.py for dictionary application and log-likelihood ratio tests (Rayson 2008); 03_robustness.py for the four robustness tests; and dict_v3.py with the dictionary itself. (ii) "data" contains the 80 original speech files in their source format, the 79 plain-text extractions after deduplication, and a manifest CSV with file-level metadata. (iii) "results" contains the reference JSON used for verification of bit-exact replication. (iv) "docs" contains the methodological log and the robustness appendix in DOCX format. The pipeline reproduces the manuscript's Table 3 figures bit-exactly when the three documented processing decisions are followed (Spanish-letter regex tokenization, MD5 deduplication of one bit-identical Uribe P1 file, and chronological reassignment of the 6 July 2006 Uribe speech to its proper subcorpus). Full instructions and a verification snippet are included in the README. Code is released under the MIT License. Corpus material is released under CC-BY 4.0; the speeches are public material produced by the offices of the President of Ecuador and the President of Colombia.