literature-search（文献检索）

Literature Search

Search multiple academic databases to find relevant papers.

Input

• $ARGUMENTS — The search query (natural language)

Scripts

Semantic Scholar (primary — best for ML/AI, has BibTeX)

python ~/.claude/skills/deep-research/scripts/search_semantic_scholar.py \
  --query "QUERY" --max-results 20 --year-range 2022-2026 \
  --api-key "$(grep S2_API_Key /Users/lingzhi/Code/keys.md 2>/dev/null | cut -d: -f2 | tr -d ' ')" \
  -o results_s2.jsonl

Key flags: --peer-reviewed-only, --top-conferences, --min-citations N, --venue NeurIPS ICML

arXiv (latest preprints)

python ~/.claude/skills/deep-research/scripts/search_arxiv.py \
  --query "QUERY" --max-results 10 -o results_arxiv.jsonl

OpenAlex (broadest coverage, free, no API key)

python ~/.claude/skills/literature-search/scripts/search_openalex.py \
  --query "QUERY" --max-results 20 --year-range 2022-2026 \
  --min-citations 5 -o results_openalex.jsonl

Merge & Deduplicate

python ~/.claude/skills/deep-research/scripts/paper_db.py merge \
  --inputs results_s2.jsonl results_arxiv.jsonl results_openalex.jsonl \
  --output merged.jsonl

CrossRef (DOI-based lookup, broadest type coverage)

python ~/.claude/skills/literature-search/scripts/search_crossref.py \
  --query "QUERY" --rows 10 --output results_crossref.jsonl

Key flags: --bibtex (output.bib format), --rows N

Download arXiv Source (get.tex files)

python ~/.claude/skills/literature-search/scripts/download_arxiv_source.py \
  --title "Paper Title" --output-dir arxiv_papers/

Key flags: --arxiv-id 1706.03762, --metadata, --max-results N

Generate BibTeX from results

python ~/.claude/skills/deep-research/scripts/bibtex_manager.py \
  --jsonl merged.jsonl --output references.bib

Workflow

1. Expand the user's query into 2-4 complementary search queries
2. Run Semantic Scholar search (primary) with expanded queries
3. Run arXiv for very recent preprints (< 3 months)
4. Optionally run OpenAlex for broader coverage
5. Merge and deduplicate results
6. Rank by: citations (0.3) + recency (0.3) + venue quality (0.2) + relevance (0.2)
7. Present structured results table

Venue Quality Tiers

Tier 1: NeurIPS, ICML, ICLR, ACL, EMNLP, NAACL, CVPR, ICCV, ECCV, KDD, AAAI, IJCAI, SIGIR, WWW Tier 2: AISTATS, UAI, COLT, COLING, EACL, WACV, JMLR, TACL Tier 3: Workshops, arXiv preprints — mark with (preprint)

Output Format

Present results as a table + detailed entries with BibTeX keys. Always note preprint status.

Related Skills

• Downstream: citation-management, literature-review, related-work-writing
• See also: deep-research, novelty-assessment