SSC Stock Comments

This repository is a page-aware database and lightweight web UI for searching North Pacific Council SSC report comments by stock, FMP, year, and comment type.

Contents

• data/processed/ssc_stock_comments.csv: page-aware comment records for analysis.
• data/processed/ssc_stock_summary.csv: stock/FMP summary counts.
• docs/: static web UI suitable for local use or GitHub Pages.
• docs/pdfs/: the unique source SSC report PDFs served by the web UI.
• scripts/build_data.py: reproducible data builder using pdftotext.

Use The UI Locally

You can open docs/index.html directly in a browser. The repository includes docs/assets/comments-data.js so local file:// use does not depend on browser permission to fetch JSON files.

From the repository root:

python3 -m http.server 8000 --directory docs

Then open:

http://localhost:8000

The Open page links point to docs/pdfs/<report>.pdf#page=<n>, which opens the source PDF near the paragraph page.

The UI links to the separate ABC/TAC app at https://jimianelli.github.io/ABC_TAC/ for ABC and OFL recommendation tables.

Publish On GitHub Pages

This repository is configured for public GitHub Pages deployment with .github/workflows/pages.yml. After pushing the repository to GitHub, set the repository Pages source to GitHub Actions under Settings -> Pages. Each push to main will publish the static app from docs/.

The public URL will be:

https://<github-user-or-org>.github.io/ssc-stock-comments/

If the repository is renamed, replace ssc-stock-comments in the URL with the GitHub repository name.

Rebuild The Data

Install Poppler if pdftotext is missing. On macOS with Homebrew:

brew install poppler

Then run:

python3 scripts/build_data.py

The script rebuilds data/processed/*.csv, docs/assets/comments.json, and docs/assets/comments-data.js.

Data Fields

• stock: normalized stock or stock complex.
• fmp: BSAI, GOA, or BSAI/GOA.
• comment_type: rule-based category such as request, recommendation, support/concur, or concern. The UI's comment type filter also includes ABC term flags from abc_buffer_terms: maximum permissible, buffer, reduction, and from maxABC.
• year, month, source_file, page: source metadata.
• page_url: UI-ready link to the PDF page.
• abc_buffer_terms: matched terminology for ABC buffers or reductions from maximum permissible ABC.
• model_choice_relationship: SSC agreement or disagreement with the assessment authors and/or Plan Team on model selection. Ambiguous candidates are unclear.
• model_choice_flag: true for explicit different-model cases and candidates marked for review.
• model_choice_confidence: high for explicit same-paragraph classifications or review when surrounding text suggests a comparison that requires verification.
• model_choice_evidence: the source sentence supporting the classification.
• excerpt: card-friendly text.
• full_text: full source paragraph.

The extraction is intentionally auditable. Broad ecosystem paragraphs may match multiple stocks when they discuss several species; use matched_terms, section, source_file, and page during curation.

Human-reviewed model-choice classifications can be added to data/model_choice_overrides.csv. Overrides are keyed by source file, page, paragraph index, and stock, and are applied whenever the data are rebuilt.