Textbook Analyzer: critical-sociopolitical analysis of PDF textbooks

This release adds a critical-sociopolitical reading module to the Textbook Analyzer. Beyond readability and topics, the tool now asks which world a textbook assumes: which places, festivals, sports, currencies, forms of work and values it treats as everyday, what it leaves out, and what that may mean for students who live somewhere else.

The module operationalises the doctoral thesis Da Cruz (2025), Beyond the pages: cultural and linguistic aspects of Canadian textbooks (UNESP), doi:10.5281/zenodo.15114000.

New

• Critical reading (step 5). Applies an editable codebook across the thesis's three dimensions (sociocultural, linguistic, sociopolitical). Shows hits in context with page numbers, density per 1,000 words, a cultural-distance indicator, absences, gendered terms and bridging suggestions. Optional LLM memo per chapter.
• Default codebook codebooks/sociopolitical_v1.json: 3 dimensions, 18 codes, terms tagged by local marker (CA, BR, AT, global) and a bridging memo per code. Can be downloaded, edited and re-uploaded in the app.
• REFI-QDA export (.qdpx) with the critical codebook and pre-coded segments, for MAXQDA, ATLAS.ti and NVivo.
• Bilingual interface (Portuguese / English).
• Project site on GitHub Pages with an in-browser version of the detector.
• Technical documentation (TECHNICAL.md): data flow, codebook format, indicators, export and limits.
• Citation metadata (CITATION.cff).

Changed

• core/llm.py: new critical-reading prompt; model names configurable via ANTHROPIC_MODEL and OPENAI_MODEL; local models via Ollama.
• core/maxqda_export.py: the experimental .mx20 package now carries the critical code tree.
• app.py: reorganised into six steps.

Infrastructure

• Automated tests with pytest (detector, exporter, app pages) running on GitHub Actions.
• Automatic deployment of the project site to GitHub Pages.

Note

Every hit is a suggestion for reading, not a finding. Content analysis remains human work (Bardin, 2016); the tool helps locate, count and move material into QDA software where the researcher codes.