CrossRef Labs

Extracto Automagically pull references out of your PDFs

What is this?

CrossRef's Extracto can extract unstructured references from article PDFs. The service wraps CrossRef Lab's PDF content extraction toolkit, pdf-extract which is an open source Ruby library currently hosted on GitHub.

How accurate is it?

The extraction toolkit is a work in progress so please don't be surprised if you get incomplete results or completely rubbish output. It's also heavily reliant on a bunch of heuristics whose defaults may not work well with your PDFs. If this service does return junk results for your PDFs you may want to try running pdf-extract with custom settings.

How reliable/fast is this service?

Not very. This is a technology preview running on development servers.

Upload a PDF for reference extraction
Optional. Used to create CrossRef deposit data for your PDF's references.
Note: Content from your PDF will be used as training data for pdf-extract. Content will not be given to third parties.
If this is not the first time you are uploading a PDF, check this box if you wish to re-process the PDF. Otherwise you will be redirected to cached results.
Upload

RESTish API

Do you have a large batch of PDFs that you'd like to submit for reference scanning programatically? This service can be used via a REST API for programmatic upload of PDFs and subsequent retrieval of references. Retrieval of references requires two or more HTTP calls - one to upload a PDF and one or more to request the status of parsed references.

Uploading a PDF

Upload a PDF by POSTing a PDF to the /pdfs resource, while also specifying a content type of application/pdf . This should not be a multi-part form post. Instead the content of the HTTP request message must be the binary content of a PDF file. For example, with curl: 

$ curl -H "Content-Type: application/pdf" --data-binary "@myfile.pdf" http://extracto.labs.crossref.org/pdfs

This request will return a JSON document containing an ID which can be used to query the status of reference parsing. 

{
  "id": "4ea837fdf8049231e4000001"
}

Retrieving references

Use the ID returned from uploading a PDF to query the status of reference parsing, and retrieve references if they are available. Sent a GET request to /pdfs/ID . 

$ curl http://extracto.labs.crossref.org/pdfs/4ea837fdf8049231e4000001

{
  "parsed": true,
  "uploaded": true,
  "failed": false,
  "file_digest": "71e176c528fcfa1b502b49465ffb51ca",
  "doi": "10.10/abc"
  "id": "4ea837fdf8049231e4000001"
}

If the status response reports that the pdf is parsed, as in data["parsed"] == true then a GET request can be made to retrieve references from /pdfs/ID/refs . If the status reports that parsing has not completed then the client should wait a few seconds and make another request for the status. 

$ curl http://extracto.labs.crossref.org/pdfs/4ea837fdf8049231e4000001/refs

[
  "Special Scrutiny: A Targeted Form of Research Protocol Review ANN INTERN MED 2004;140:220-223",
  ...
]

Generating CrossRef deposit XML

Extracto can produce CrossRef deposit XML for parsed PDFs. Deposit XML including parsed references for a particular PDF will be generated and returned on a GET request to /pdfs/ID/deposit .