PubScholar
This section describes how the OpenAIRE Graph integrates the metadata of the journal articles published by PubScholar, a Chinese platform for scholarly resources. It covers where the data comes from, how it is acquired and transformed, and how the PubScholar metadata fields populate the fields of the OpenAIRE Graph data model.
Data source
PubScholar is a public academic platform built by the Chinese Academy of Sciences (CAS) and operated by the National Science Library of CAS, launched in November 2023. It provides discovery of and access to scholarly resources: journal articles, books, institutional repositories, theses, ChinaXiv preprints, scientific data and patents.
PubScholar shares with OpenAIRE the metadata of the articles of Chinese scholarly journals. The journals are those of a list agreed with PubScholar, which currently contains 213 journals.
Data acquisition
The metadata is collected from the PubScholar Article Retrieval API, which returns the articles of one journal at a time as JSON documents. OpenAIRE harvests the articles of every journal in the list, removes the duplicates (articles returned for more than one journal), and keeps one JSON document per article. The harvest of June 2026 contained 1,015,037 articles.
Every collection imports a complete harvest, so the PubScholar records in the Graph reflect the harvest available at collection time.
The API returns, for each article: identifier, title, abstract, authors with their institutions, keywords, DOI, link to the landing page, journal name, volume, issue, pages, publication year, and the indexing databases that cover the journal. It provides no access-right, language or publisher information.
Processing workflow
- Harvest: the articles of the journals in the list are retrieved from the PubScholar API.
- Conversion: every JSON document becomes one XML record. The conversion only changes the syntax: each JSON field becomes an XML element with the same value, and each list becomes a repeated element.
The record header carries the PubScholar identifier of the article (as
pubscholar:<id>), the harvest date and the ISSN of the journal, taken from the journal list. Empty values are dropped, and the few characters that XML 1.0 does not allow are removed. - Collection: the OpenAIRE aggregation system collects the resulting file for the data source PubScholar, one record per article.
- Transformation: the transformation rule
xslt_cleaning_pubscholar_hadoopmaps each record into the OpenAIRE internal metadata format, which is based on the DataCite schema. The mapping choices described below are made at this step. - Graph production: each transformed record becomes a research product, which then goes through the standard steps of the Graph production workflow (cleaning, deduplication, enrichment).
Record filtering
A record is not transformed when it has neither a doi nor a link (about 0.85% of the articles).
Entity Mapping
The table below describes how the fields of a PubScholar article populate the fields of an OpenAIRE research product.
JSON paths refer to the article document returned by the API; [] marks a list.
| OpenAIRE Research Product field path | PubScholar record JSON path | Notes |
|---|---|---|
| Research product | ||
id | id | id in the form 735bb412454f::md5(pubscholar:<id>), built from the local identifier, because PubScholar is not a PID authority |
originalIds | id | pubscholar:<id> and the internal identifier |
type | (fixed value) | publication: the API returns journal articles only |
mainTitle | title | |
authors | authors[] | One author per name, in the order of the source; see Authors |
publicationDate | year | See Publication dates |
language | title, abstracts | The source has no language field; see Language |
subjects | keywords[] | scheme=keyword |
bestAccessRight | Derived from instances.accessRight. Later steps of the Graph production can set it to OPEN, for example for articles of gold open access journals (openAccessColor gold) | |
collectedfrom | (fixed value) | The data source PubScholar, openaire____::5b0a44e9440d46c9597500f1bbfd2643. The same data source is set as the host of the instance |
| Author Mapping | ||
authors.fullName | authors[] | The name as given by the source, not split into given name and surname |
authors.rank | Sequential number starting from 1 | |
| Container Mapping | ||
container.name | journal | |
container.issnPrinted | (journal list) | The ISSN of the journal in the list used for the harvest; the article itself carries no ISSN |
container.vol | volume | |
container.iss | issue | |
container.sp | first_page | |
container.ep | last_page | |
| Instance Mapping | One instance per record | |
instances.type | (fixed value) | Article, from the dnet:publication_resource vocabulary |
instances.alternateIdentifiers | doi | The DOI, when present (about 45% of the articles), with scheme = doi. PubScholar is not a PID authority for DOIs, so the DOI is an alternate identifier, not a value of instances.pids |
instances.urls | link, doi | The landing page link, unless it is the DOI URL, and the DOI URL https://doi.org/<doi> when a DOI is present. For articles without a DOI, the landing page is the only locator of the record |
instances.accessRight | (fixed value) | UNKNOWN: the source has no access-right information, and the landing pages are often behind a paywall |
instances.publicationDate | year | As publicationDate |
instances.refereed | (fixed value) | peerReviewed: the journals in the list are peer-reviewed journals |
All the records are related to the data source with name=PubScholar and id=openaire____::5b0a44e9440d46c9597500f1bbfd2643.
Authors
Each name in authors[] becomes an author, with the institutions listed for that name in relations[] (relations[].name equal to the author name, relations[].institution[]) kept as raw affiliation strings.
The institutions carry no identifier, so they do not create links to organizations.
Language
The source has no language field. The language is set to Chinese when the title or the abstract contains Chinese characters, and to English otherwise; it is then cleaned by using vocabulary dnet:languages.
Publication dates
PubScholar gives only the publication year, which becomes publicationDate as the first day of that year.
A year is accepted only between 1300 and the next calendar year; outside this window the research product has no publication date.
Relation Mapping
The PubScholar records do not create relations: the source has no references, no funding information, and no identifiers for the institutions of the authors.
Fields not mapped
The following PubScholar fields do not populate the Graph: the abstract (abstracts, used only to detect the language), the indexing databases that cover the journal (collections[]), and the aggregated list of institutions (institutions[], which repeats the institutions of the authors).