openalex-paper-search

学术论文搜索（OpenAlex）

通过 OpenAlex API 搜索 2.4 亿多篇学术作品——完全免费，无需 API 密钥，也无需 SDK。只需使用 curl 或 bash 构造 URL。

完整文档： https://docs.openalex.org

快速开始

OpenAlex 是一个 REST API。通过构造 URL 并使用 curl 获取数据来发起查询。所有响应均为 JSON。

# Search for papers about "transformer architecture"
curl -s "https://api.openalex.org/works?search=transformer+architecture&per_page=5&[email protected]" | python3 -m json.tool

重要提示： 每次请求都必须包含 [email protected]（或任何有效的电子邮件地址）。否则，每秒只能发送 1 个请求。包含该参数后，每秒可发送 10 个请求（“礼貌池”）。

核心概念

实体

OpenAlex 包含以下实体类型（均可查询）：

实体	端点	数量	描述
作品	/works	2.4 亿+	论文、文章、书籍、数据集、学位论文
作者	/authors	9000 万+	创作作品的人
来源	/sources	25 万+	期刊、知识库、会议
机构	/institutions	11 万+	大学、研究机构
主题	/topics	4000+	研究主题（分层级）

作品对象——关键字段

获取作品时，以下字段最实用：

id                        OpenAlex ID (e.g., "https://openalex.org/W2741809807")
doi                       DOI URL
title / display_name      Paper title
publication_year          Year published
publication_date          Full date (YYYY-MM-DD)
cited_by_count            Number of incoming citations
fwci                      Field-Weighted Citation Impact (normalized)
type                      article, preprint, review, book, dataset, etc.
language                  ISO 639-1 code (e.g., "en")
is_retracted              Boolean
open_access.is_oa         Boolean -- is it freely accessible?
open_access.oa_url        Direct URL to free version
authorships               List of authors with names, institutions, ORCIDs
abstract_inverted_index   Abstract as inverted index (needs reconstruction)
referenced_works          List of OpenAlex IDs this work cites (outgoing)
related_works             Algorithmically related works
cited_by_api_url          API URL to get works that cite this one (incoming)
topics                    Assigned research topics with scores
keywords                  Extracted keywords with scores
primary_location          Where the work is published (journal, repo)
best_oa_location          Best open access location with PDF link

重构摘要

出于法律原因，OpenAlex 将摘要存储为倒排索引。要获取纯文本，请按以下方式重构：

import json, sys
# Read the abstract_inverted_index from a work object
inv_idx = work["abstract_inverted_index"]
if inv_idx:
    words = [""] * (max(max(positions) for positions in inv_idx.values()) + 1)
    for word, positions in inv_idx.items():
        for pos in positions:
            words[pos] = word
    abstract = " ".join(words)

也可以在 bash 中使用 python3 -c：

# Pipe a work JSON into this to extract the abstract
echo "$WORK_JSON" | python3 -c "
import json,sys
w=json.load(sys.stdin)
idx=w.get('abstract_inverted_index',{})
if idx:
    words=['']*( max(max(p) for p in idx.values())+1 )
    for word,positions in idx.items():
        for pos in positions: words[pos]=word
    print(' '.join(words))
"

搜索论文

基本关键词搜索

搜索标题、摘要和全文。支持词干提取和停用词移除。

# Simple search
curl -s "https://api.openalex.org/works?search=large+language+models&[email protected]"

# With per_page limit
curl -s "https://api.openalex.org/works?search=CRISPR+gene+editing&per_page=10&[email protected]"

布尔搜索

使用大写的 AND、OR、NOT，并配合括号和带引号的短语：

# Complex boolean query
curl -s "https://api.openalex.org/works?search=(reinforcement+learning+AND+%22robot+control%22)+NOT+simulation&[email protected]"

# Exact phrase match (use double quotes, URL-encoded as %22)
curl -s "https://api.openalex.org/works?search=%22attention+is+all+you+need%22&[email protected]"

搜索特定字段

# Title only
curl -s "https://api.openalex.org/works?filter=title.search:transformer&[email protected]"

# Abstract only
curl -s "https://api.openalex.org/works?filter=abstract.search:protein+folding&[email protected]"

# Title and abstract combined
curl -s "https://api.openalex.org/works?filter=title_and_abstract.search:neural+scaling+laws&[email protected]"

# Fulltext search (subset of works)
curl -s "https://api.openalex.org/works?filter=fulltext.search:climate+tipping+points&[email protected]"

筛选

筛选器是最强大的功能。使用逗号（AND）或竖线（OR）组合筛选条件。

最实用的筛选器

# By publication year
?filter=publication_year:2024
?filter=publication_year:2020-2024
?filter=publication_year:>2022

# By citation count
?filter=cited_by_count:>100        # highly cited
?filter=cited_by_count:>1000       # landmark papers

# By open access
?filter=is_oa:true                 # only open access
?filter=oa_status:gold             # gold OA only

# By type
?filter=type:article               # journal articles
?filter=type:preprint              # preprints
?filter=type:review                # review articles

# By language
?filter=language:en                # English only

# Not retracted
?filter=is_retracted:false

# Has abstract
?filter=has_abstract:true

# Has downloadable PDF
?filter=has_content.pdf:true

# By author (OpenAlex ID)
?filter=author.id:A5023888391

# By institution (OpenAlex ID)
?filter=institutions.id:I27837315  # e.g., University of Michigan

# By DOI
?filter=doi:https://doi.org/10.1038/s41586-021-03819-2

# By indexed source
?filter=indexed_in:arxiv           # arXiv papers
?filter=indexed_in:pubmed          # PubMed papers
?filter=indexed_in:crossref        # Crossref papers

组合筛选器

# AND: comma-separated
?filter=publication_year:>2022,cited_by_count:>50,is_oa:true,type:article

# OR: pipe-separated within a filter
?filter=publication_year:2023|2024

# NOT: prefix with !
?filter=type:!preprint

# Combined example: highly-cited OA articles from 2023-2024, not preprints
curl -s "https://api.openalex.org/works?filter=publication_year:2023-2024,cited_by_count:>50,is_oa:true,type:!preprint&search=machine+learning&per_page=10&[email protected]"

排序

# Most cited first
?sort=cited_by_count:desc

# Most recent first
?sort=publication_date:desc

# Most relevant first (only when using search)
?sort=relevance_score:desc

# Multiple sort keys
?sort=publication_year:desc,cited_by_count:desc

分页

有两种模式：基本分页（用于浏览）和游标分页（用于收集所有结果）。

# Basic paging (limited to 10,000 results)
?page=1&per_page=25
?page=2&per_page=25

# Cursor paging (unlimited, for collecting everything)
?per_page=100&cursor=*                    # first page
?per_page=100&cursor=IlsxNjk0ODc...      # next page (cursor from previous response meta)

下一页的游标位于 response.meta.next_cursor 中。当它为 null 时，表示已到末尾。

选择字段

只选择所需字段，以减小响应体积：

# Only get IDs, titles, citation counts, and DOIs
?select=id,display_name,cited_by_count,doi,publication_year

# Minimal metadata for scanning
?select=id,display_name,publication_year,cited_by_count,open_access

引用图遍历

查找某篇论文引用的内容（出向引用）

# Get works cited BY a specific paper
curl -s "https://api.openalex.org/works?filter=cited_by:W2741809807&per_page=25&[email protected]"

查找引用某篇论文的内容（入向引用）

# Get works that CITE a specific paper
curl -s "https://api.openalex.org/works?filter=cites:W2741809807&sort=cited_by_count:desc&per_page=25&[email protected]"

查找相关作品

# Get related works (algorithmic, based on shared concepts)
curl -s "https://api.openalex.org/works?filter=related_to:W2741809807&per_page=25&[email protected]"

引用链：沿引用关系追踪

1. 通过 DOI 获取一篇开创性论文
2. 查找其 referenced_works（即它引用的作品）
3. 查找引用它的作品（filter=cites:WORK_ID）
4. 对引用量最高的作品，重复上述步骤

这就是围绕某个主题构建文献图谱的方法。

作者查询

# Search for an author
curl -s "https://api.openalex.org/authors?search=Yann+LeCun&[email protected]"

# Get an author's works (by OpenAlex author ID)
curl -s "https://api.openalex.org/works?filter=author.id:A5064850633&sort=cited_by_count:desc&per_page=10&[email protected]"

# Get an author by ORCID
curl -s "https://api.openalex.org/authors/orcid:[email protected]"

按外部 ID 查询

# By DOI
curl -s "https://api.openalex.org/works/doi:10.1038/[email protected]"

# By PubMed ID
curl -s "https://api.openalex.org/works/pmid:[email protected]"

# By arXiv ID (via DOI)
curl -s "https://api.openalex.org/works/doi:10.48550/[email protected]"

# Batch lookup: up to 50 IDs at once
curl -s "https://api.openalex.org/works?filter=doi:https://doi.org/10.1234/a|https://doi.org/10.1234/b|https://doi.org/10.1234/c&[email protected]"

开放获取与 PDF 访问

# Find OA papers with direct PDF links
curl -s "https://api.openalex.org/works?search=quantum+computing&filter=is_oa:true,has_content.pdf:true&select=id,display_name,open_access,best_oa_location&per_page=5&[email protected]"

best_oa_location.pdf_url 字段会在有可用链接时提供 PDF 直链。open_access.oa_url 则提供最佳的开放获取落地页或 PDF。

实用工作流

某个主题的文献综述

# 1. Find the most-cited papers on a topic
curl -s "https://api.openalex.org/works?search=retrieval+augmented+generation&sort=cited_by_count:desc&filter=publication_year:>2020,type:article,has_abstract:true&per_page=20&select=id,display_name,publication_year,cited_by_count,doi,authorships,abstract_inverted_index&[email protected]"

# 2. For the top papers, explore their citation graphs
curl -s "https://api.openalex.org/works?filter=cites:W4285719527&sort=cited_by_count:desc&per_page=10&select=id,display_name,publication_year,cited_by_count,doi&[email protected]"

# 3. Find recent papers building on this work
curl -s "https://api.openalex.org/works?filter=cites:W4285719527,publication_year:>2023&sort=publication_date:desc&per_page=10&[email protected]"

查找里程碑式／开创性论文

# Highly cited + search term
curl -s "https://api.openalex.org/works?search=attention+mechanism+neural+networks&filter=cited_by_count:>500,type:article&sort=cited_by_count:desc&per_page=10&select=id,display_name,publication_year,cited_by_count,doi&[email protected]"

查找近期预印本

# Latest preprints on a topic
curl -s "https://api.openalex.org/works?search=multimodal+large+language+models&filter=type:preprint,publication_year:2025&sort=publication_date:desc&per_page=15&[email protected]"

查找综述文章

# Review/survey papers on a topic
curl -s "https://api.openalex.org/works?search=federated+learning&filter=type:review,cited_by_count:>20&sort=cited_by_count:desc&per_page=10&[email protected]"

作者分析

# 1. Find the author
curl -s "https://api.openalex.org/authors?search=Geoffrey+Hinton&select=id,display_name,works_count,cited_by_count,last_known_institutions&[email protected]"

# 2. Get their most influential papers
curl -s "https://api.openalex.org/works?filter=author.id:A5068082743&sort=cited_by_count:desc&per_page=10&select=id,display_name,publication_year,cited_by_count,doi&[email protected]"

# 3. Get their recent work
curl -s "https://api.openalex.org/works?filter=author.id:A5068082743,publication_year:>2023&sort=publication_date:desc&per_page=10&[email protected]"

将结果保存到磁盘

开展深入研究时，将论文数据保存到磁盘，以便后续处理：

# Save search results as JSON
curl -s "https://api.openalex.org/works?search=topic&per_page=50&[email protected]" > research/papers/topic-search.json

# Extract and save a clean summary
curl -s "https://api.openalex.org/works?search=topic&per_page=50&select=id,display_name,publication_year,cited_by_count,doi,authorships&[email protected]" | python3 -c "
import json, sys
data = json.load(sys.stdin)
for w in data.get('results', []):
    authors = ', '.join(a['author']['display_name'] for a in w.get('authorships', [])[:3])
    if len(w.get('authorships', [])) > 3: authors += ' et al.'
    print(f\"[{w.get('cited_by_count',0)} cites] {w['display_name']} ({w.get('publication_year','?')}) - {authors}\")
    if w.get('doi'): print(f\"  DOI: {w['doi']}\")
    print()
" > research/papers/topic-summary.txt

开展深入研究时，将单篇论文的元数据保存到 sources-index.md，并将原始数据保存到 sources/：

# Save a paper's full metadata
curl -s "https://api.openalex.org/works/[email protected]" > research/sources/001-paper-title.json

速率限制

池	速率	获取方式
普通	1 个请求/秒	未提供电子邮件地址
礼貌	10 个请求/秒	在请求中添加 [email protected]
高级	更高	通过 api_key 参数使用付费 API 密钥

始终使用礼貌池。 在每个请求中添加 &[email protected]。

提示

• 积极使用 select，以减少响应大小并加快请求速度
• 收集大量结果时使用 per_page=100（最大值），以尽量减少请求次数
• 需要超过 10,000 条结果时使用游标分页（cursor=*）
• 使用 OR 语法批量查询 DOI：filter=doi:DOI1|DOI2|DOI3（最多 50 个）
• 使用倒排索引重建摘要——不要跳过这一步，摘要非常有价值
• 沿引用链查找具有开创性的研究和最新进展
• 需要摘要时按 has_abstract:true 筛选（并非所有作品都有摘要）
• 按 indexed_in:arxiv 或 indexed_in:pubmed 筛选，以定位特定存储库
• 按 cited_by_count:desc 排序，优先查找影响力最大的论文
• 结合搜索和筛选以获得精准结果：搜索提供相关性，筛选提升精确度