SuperAPI AI Directory
Partner with us中文AI newsFavorites
Skillspaper-research

paper-research

Writing & research

View repository

Updated 10-16 09:53

About this skill

Paper Research — Academic Literature Search Tool

Search multiple academic databases in parallel via the command line, automatically deduplicate and merge, download PDFs, and export BibTeX.

Prerequisites

  1. CLI binary available (install to ~/.local/bin/ or any directory in $PATH):

    scholar-cli --help

  2. For WebBridge sources (Google Scholar, CNKI, WoS), kimi-webbridge daemon must be running:

    ~/.kimi-webbridge/bin/kimi-webbridge status

    If not running: invoke the kimi-webbridge skill.

Command Quick Reference

# 英文多源并行搜索（默认 7 个源）
scholar-cli search-en --query "关键词" --limit 10

# 指定源搜索
scholar-cli search-en --query "关键词" --sources openalex,semantic,arxiv,dblp

# Google Scholar（需要 WebBridge + 浏览器）
scholar-cli search-google --query "关键词" --limit 10

# 知网搜索（需要 WebBridge + 知网登录）
scholar-cli search-cnki --query "关键词" --limit 20

# Web of Science（需要 WebBridge + 机构登录）
scholar-cli search-wos --query "关键词" --limit 10

# 按 DOI 查论文详情
scholar-cli detail --doi "10.1145/3442188.3445922"

# 搜索并保存到工作区（自动去重）
scholar-cli search-en --query "关键词" --workspace /path/to/workspace

# 导出 BibTeX
scholar-cli export --workspace /path/to/workspace --output refs.bib

# 检查 WoS 登录状态
scholar-cli login-status --platform wos

All Available Sources

Source Source Name Type Login Required
OpenAlex openalex HTTP API No
Semantic Scholar semantic HTTP API No
CrossRef crossref HTTP API No
arXiv arxiv HTTP API No
PubMed pubmed HTTP API No
DBLP dblp HTTP API No
bioRxiv/medRxiv biorxiv HTTP API No
Google Scholar — WebBridge No (but has CAPTCHA)
CNKI — WebBridge Yes
Web of Science — WebBridge Yes (institutional)

search-en By default, searches the first 7 HTTP API sources in parallel. Google Scholar, CNKI, and WoS require separate commands.

Field → Source Mapping

The core literature distribution differs greatly across disciplines; choosing the right sources directly determines search quality.

Field Preferred Sources Secondary Sources Notes
Computer Science / AI / ML arxiv, semantic, dblp openalex arXiv is the main venue for CS preprints, S2 has the best CS coverage, and DBLP is the authoritative CS bibliography
Biomedical / Clinical Medicine pubmed, biorxiv crossref, openalex PubMed is a must-check entry point to MEDLINE; bioRxiv/medRxiv cover biomedical preprints
Physics / Mathematics / Astronomy arxiv openalex, crossref Physics and mathematics rely heavily on arXiv
Social Sciences / Economics openalex, crossref semantic Mainly traditional journals; OpenAlex has the broadest coverage
Engineering / Materials / Chemistry crossref, openalex semantic Mainly traditional journals; CrossRef has the most complete DOI metadata
Chinese-language literature (any discipline) cnki(CNKI) — The only reliable source for Chinese journals, dissertations, and conference papers. Login required
High-impact/citation analysis search-wos openalex, semantic WoS has the most authoritative citation data, suitable for citation analysis and finding high-impact papers
Interdisciplinary / review research All sources search-en search-google When the field is uncertain, run all sources in parallel, then look at the distribution after deduplication and focus

Usage Strategy

  1. Known field → use --sources to specify 2-3 core sources and reduce noise

    # AI / CS 论文
scholar-cli search-en --query "large language model reasoning" --sources arxiv,semantic,dblp --limit 15

# 生物医学论文
scholar-cli search-en --query "CRISPR gene therapy" --sources pubmed,biorxiv,crossref --limit 15

# 工程/材料
scholar-cli search-en --query "perovskite solar cell" --sources crossref,openalex --limit 15

  2. Uncertain field → first run a search across all sources, then look at the sources summary to determine which sources have material

    scholar-cli search-en --query "quantum computing" --limit 5
# 看返回的 sources 字段，哪个 count 高就追加搜哪个

  3. Chinese research → use search-cnkiand ensure the browser is logged in to CNKI

    scholar-cli search-cnki --query "大语言模型教育应用" --limit 20

  4. Citation analysis/high-impact filtering → use WoS; institutional VPN/login required

    scholar-cli search-wos --query "transformer attention" --limit 20

  5. Need Google Scholar → when blocked by CAPTCHA, have the user manually solve the CAPTCHA in the browser; then it will be available for a short time

    scholar-cli search-google --query "attention mechanism survey"

Search Keyword Tips

  • English sources: use English keywords, supports Boolean logic (depending on each source's API)
  • CNKI: use Chinese keywords, consistent with CNKI web search behavior
  • DBLP: especially accurate for matching CS conference/journal paper titles, suitable for precise lookup
  • bioRxiv: Implemented via CrossRef DOI prefix filtering, search quality is consistent with CrossRef
  • DOI detail lookup: If you got a DOI from the search results, use the detail command to complete the full information (author affiliations, abstract, citation count, PDF link)

Workspace management

The workspace is used to accumulate papers across multiple searches, automatically deduplicated by DOI / title+authors:

# 多次搜索存同一个工作区
scholar-cli search-en --query "term1" --workspace ~/research/my-survey --limit 10
scholar-cli search-en --query "term2" --workspace ~/research/my-survey --limit 10

# 最后统一导出 BibTeX
scholar-cli export --workspace ~/research/my-survey --output ~/research/my-survey/refs.bib

In the return value, papers_added indicates how many new items were added (existing ones are not added again),total_stored is the workspace total.

Output format

All commands output JSON:

  • Success:{"ok": true, "data": {...}}
  • Failure:{"ok": false, "error": {"code": "...", "message": "..."}}

Paper fields: title, authors, abstract, year, doi, venue, volume, issue, pages, citations, references, open_access, pdf_url, source, sources, urls, identifiers

FAQ

Issue Cause and solution
PubMed returns "blocked by NCBI" NCBI has rate-limited the requests or IP; wait a while and retry or add an API key
DBLP connection timeout Access to dblp.org is unstable in some network environments; does not affect other sources
Google Scholar CAPTCHA Just manually pass the captcha once in the browser
CNKI "not_logged_in" Open cnki.net in the browser and log in (see the CNKI usage instructions below)
CNKI slider captcha It appears not only at login; new sessions, frequent requests, and automation detection all trigger it (see the CNKI usage instructions below)
WoS "not_logged_in" Requires logging in to Web of Science via institutional VPN or SSO
Semantic Scholar returns 0 results S2's search is sensitive to keywords; try a more specific query
bioRxiv year is 0 Very few early papers lack date metadata
A certain source times out Does not affect other sources; sources in the results will show error information

CNKI usage instructions

Before using search-cnki , you must complete the following preparations in the browser:

  1. Log in to CNKI — Open https://www.cnki.net in Chrome and log in to your account
  2. Pass the slider captcha once — Open https://kns.cnki.net and search any keyword; if the slider captcha pops up (the "security verification" page), manually drag the slider to complete the verification

CNKI's slider captcha (blockPuzzle type) appears not only at login; it is also triggered in the following scenarios:

  • First visit from a new session / new IP
  • Too many searches within a short time
  • Automated browser behavior detected

The captcha cannot be cracked automatically, and must be completed manually. After passing it once, the same browser session can be used normally for a period of time.

If search-cnki returns an error or empty results, first have the user visit kns.cnki.net in the browser to confirm whether the captcha needs to be passed again.

Dependencies

  • 7 HTTP API sources（OpenAlex, Semantic Scholar, CrossRef, arXiv, PubMed, DBLP, bioRxiv）：No login required, direct HTTP calls
  • Google Scholar: requires kimi-webbridge running + Chrome open
  • CNKI: requires kimi-webbridge running + Chrome open + CNKI logged in + Manually pass the slider CAPTCHA once
  • Web of Science: Requires kimi-webbridge running + Chrome open + institution SSO login