| name | fetch-awesome-repo-rec-sys-paper |
| description | Resolves paper titles and repo-relative PDF paths from the curated Awesome-Deep-Learning-Papers-for-Search-Recommendation-Advertising index by keyword. Returns section, title, decoded path inside the companion PDF repo, and the GitHub blob URL.
Triggers: user names a paper / acronym / author / venue and wants the canonical path or URL from the Awesome rec-sys list (e.g. "find DIEN", "where is the wide & deep paper", "Awesome rec-sys list").
|
Fetch paper path from Awesome rec-sys list
Looks up papers from the curated Awesome-Deep-Learning-Papers-for-Search-Recommendation-Advertising README. The README only stores markdown links; PDFs live in the companion repo guyulongcs/Deep-Learning-for-Search-Recommendation-Advertisements. The repo-relative path is the segment after blob/<branch>/ in each link, URL-decoded.
Inputs
| Input | Description | Example |
|---|
| Query tokens | One or more space-separated tokens; every token must appear in title / path / URL | DIEN, wide deep, multi-task ranking |
| README path (optional) | Override via env AWESOME_REC_SYS_README or --readme flag | — |
Tokens that are alphanumeric/hyphen are matched as whole words (so DIEN does not match Gradient).
References (load when relevant)
- None — this skill is self-contained.
Steps
-
Locate the README. Default lookup order in the script:
AWESOME_REC_SYS_README env var
- Hard-coded fallback at
/path/to/works/for/you/Projects/Awesome-Deep-Learning-Papers-for-Search-Recommendation-Advertising/README.md
If neither resolves, ask the user to clone the Awesome repo or set the env var.
-
Run the script:
python3 ~/.claude/skills/fetch-awesome-repo-rec-sys-paper/scripts/find_paper.py "DIEN"
python3 ~/.claude/skills/fetch-awesome-repo-rec-sys-paper/scripts/find_paper.py wide deep
-
Return the matches — section heading, human title, repo-relative path: (decoded), and full GitHub url:. Report all matches when duplicates appear under different sections (e.g. same paper under 03_Ranking and Sequence-Modeling).
Fallback without the script
- Open the README at the resolved path.
- Search case-insensitive for the keyword.
- For each
* [Title](URL) line with a .pdf URL, decode the path after blob/master/ (e.g. %20 → space).
- Return section (nearest preceding
## heading), title, relative path, and full URL.
Notes
- Link titles use nested brackets (e.g.
[DIN] inside [...](url)) — the script handles this. A naive [^]]+ regex on the whole line will fail.
- Markdown emphasis (
**, _, backticks) in link text is stripped for display.
- Local PDFs only exist if the user has separately cloned
Deep-Learning-for-Search-Recommendation-Advertisements; the Awesome README still gives the correct relative path.
Output
- One block per match:
section, title, path: (relative path inside the PDF repo), url: (full GitHub blob URL).