openaire-graph-docs/docs/data-provision/aggregation/pubmed.md

44 lines
12 KiB
Markdown
Raw Normal View History

# PubMed
This section describes the mapping implemented for [MEDLINE/PubMed](https://pubmed.ncbi.nlm.nih.gov/).
## Input
The native data is collected from the [ftp baseline](https://ftp.ncbi.nlm.nih.gov/pubmed/baseline/) site.
It contains XML records compliant with the schema available at [www.nlm.nih.gov](https://www.nlm.nih.gov/bsd/licensee/elements_descriptions.html).
## Incremental harvesting
Pubmed exposes an entry point FTP with all the updates for each one. [ftp baseline update](https://ftp.ncbi.nlm.nih.gov/pubmed/updatefiles/). We collect the new file and generate the new dataset by upserting the existing item.
2022-11-08 09:35:10 +01:00
## Entity Mapping
The table below describes the mapping from the XML baseline records to the OpenAIRE Graph dump format.
| OpenAIRE Result field path | PubMed record field xpath | Notes |
|--------------------------------|-----------------------------------------------------------------------------------------------------------------------------------------------|-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| **Publication Mapping** | | |
| `id` | `//PMID` | id in the form `pmid_________::md5(pmid)` |
| `pid` | `//PMID` | `classid = classname = pmid` |
| `publicationdate` | `//PubmedPubDate` | clean and normalize the format of the date to be YYYY-mm-dd |
| `maintitle` | `//Title` | |
| `description` | `//AbstractText` | |
| `language` | `//Language` | cleaning vocabulary -> dnet:languages |
| `subjects` | `//DescriptorName` | classId, className = keyword |
| **Author Mapping** | | |
| `author.surname` | `//Author/LastName` | |
| `author.name` | `//Author/ForeName` | |
| `author.fullname` | `//Author/FullName` | Concatenation of forename + lastName if exist |
| `author.rank` | FOR ALL AUTHORS | sequential number starting from 1 |
| **Journal Mapping** | | |
| `container.conferencedate` | `//Journal/PubDate` | map the date of the Journal |
| `container.name` | `//Journal/Title` | name of the journal |
| `container.vol` | `//Journal/Volume` | journal volume |
| `container.issPrinted` | `//Journal/ISSN` | the journal issn |
| `container.iss` | `//Journal/Issue` | The journal issue |
| **Instance Mapping** | | |
| `instance.type` | `//PublicationType` | if the article contains the typology `Journal Article` then we apply this type else We have to find a terms that match the vocabulary otherwise we discard it |
| `type` | <ul><li>`\attributes\types\resourceType`</li> <li> `\attributes\types\resourceTypeGeneral` </li> <li>`attributes\types\schemaOrg`</li></ul> | Using the **_dnet:result_typologies_** vocabulary, we look up the `instance.type` synonym to generate one of the following main entities: <ul><li>`publication`</li> <li>`dataset`</li> <li> `software`</li> <li>`otherresearchproduct`</li></ul> |
| `instance.pid` | `//PMID` | map the pmid in the pid in the instance |
| `instance.url` | `//PMID` | creates the URL by prepending `https://pubmed.ncbi.nlm.nih.gov/` to the PMId |
| `instance.alternateIdentifier` | `//ArticleId[./@IdType="doi"]` | |
| `instance.publicationdate` | `//PubmedPubDate` | clean and normalize the format of the date to be YYYY-mm-dd |