233 lines
9.0 KiB
Python
233 lines
9.0 KiB
Python
"""
|
|
This files contains all the functions to extract DOIs of citations from
|
|
PDF files.
|
|
|
|
# TODO: Unittests
|
|
"""
|
|
import os
|
|
import requests
|
|
import subprocess
|
|
import xml.etree.ElementTree as ET
|
|
|
|
from requests.exceptions import RequestException
|
|
|
|
from libbmc import tools
|
|
from libbmc.citations import plaintext
|
|
|
|
|
|
CERMINE_BASE_URL = "http://cermine.ceon.pl/"
|
|
SCRIPT_DIR = os.path.dirname(os.path.realpath(__file__))
|
|
|
|
|
|
def cermine(pdf_file, force_API=False, override_local=None):
|
|
"""
|
|
Run `CERMINE <https://github.com/CeON/CERMINE>`_ to extract metadata from \
|
|
the given PDF file, to retrieve citations (and more) from the \
|
|
provided PDF file. This function returns the raw output of \
|
|
CERMINE call.
|
|
|
|
.. note::
|
|
|
|
Try to use a local CERMINE JAR file, and falls back to using the API. \
|
|
JAR file is expected to be found in \
|
|
``libbmc/external/cermine.jar``. You can override this using \
|
|
the ``override_local`` parameter.
|
|
|
|
.. note::
|
|
|
|
CERMINE JAR file can be found at \
|
|
`<http://maven.icm.edu.pl/artifactory/simple/kdd-releases/pl/edu/icm/cermine/cermine-impl/>`_.
|
|
|
|
.. note::
|
|
|
|
This fallback using the \
|
|
`CERMINE API <http://cermine.ceon.pl/about.html>`_, and \
|
|
hence, uploads the PDF file (so uses network). Check out \
|
|
the CERMINE API terms.
|
|
|
|
:param pdf_file: Path to the PDF file to handle.
|
|
:param force_API: Force the use of the Cermine API \
|
|
(and do not try to use a local JAR file). Defaults to ``False``.
|
|
:param override_local: Use this specific JAR file, instead of the one at \
|
|
the default location (``libbmc/external/cermine.jar``).
|
|
:returns: Raw output from CERMINE API or ``None`` if an error occurred. \
|
|
No post-processing is done.
|
|
"""
|
|
try:
|
|
# Check if we want to load the local JAR from a specific path
|
|
local = override_local
|
|
# Else, try to stat the JAR file at the expected local path
|
|
if (local is None) and (not force_API):
|
|
if os.path.isfile(os.path.join(SCRIPT_DIR,
|
|
"../external/cermine.jar")):
|
|
local = os.path.join(SCRIPT_DIR,
|
|
"../external/cermine.jar")
|
|
|
|
# If we want to force the API use, or we could not get a local JAR
|
|
if force_API or (local is None):
|
|
print("Using API")
|
|
with open(pdf_file, "rb") as fh:
|
|
# Query the API
|
|
r = requests.post(
|
|
CERMINE_BASE_URL + "extract.do",
|
|
headers={"Content-Type": "application/binary"},
|
|
files={"file": fh}
|
|
)
|
|
return r.text
|
|
# Else, use the local JAR file
|
|
else:
|
|
return subprocess.check_output([
|
|
"java",
|
|
"-cp", local,
|
|
"pl.edu.icm.cermine.PdfNLMContentExtractor",
|
|
"-path", pdf_file]).decode("utf-8")
|
|
except (RequestException,
|
|
subprocess.CalledProcessError,
|
|
FileNotFoundError):
|
|
# In case of any error, return None
|
|
return None
|
|
|
|
|
|
def cermine_dois(pdf_file, force_API=False, override_local=None):
|
|
"""
|
|
Run `CERMINE <https://github.com/CeON/CERMINE>`_ to extract DOIs of cited \
|
|
papers from a PDF file.
|
|
|
|
.. note::
|
|
|
|
Try to use a local CERMINE JAR file, and falls back to using the API. \
|
|
JAR file is expected to be found in \
|
|
``libbmc/external/cermine.jar``. You can override this using \
|
|
the ``override_local`` parameter.
|
|
|
|
.. note::
|
|
|
|
CERMINE JAR file can be found at \
|
|
`<http://maven.icm.edu.pl/artifactory/simple/kdd-releases/pl/edu/icm/cermine/cermine-impl/>`_.
|
|
|
|
.. note::
|
|
|
|
This fallback using the \
|
|
`CERMINE API <http://cermine.ceon.pl/about.html>`_, and \
|
|
hence, uploads the PDF file (so uses network). Check out \
|
|
the CERMINE API terms.
|
|
|
|
.. note::
|
|
|
|
This function uses CERMINE to extract references from the paper, and \
|
|
try to match them on Crossref to get DOIs.
|
|
|
|
:param pdf_file: Path to the PDF file to handle.
|
|
:param force_API: Force the use of the Cermine API \
|
|
(and do not try to use a local JAR file). Defaults to ``False``.
|
|
:param override_local: Use this specific JAR file, instead of the one at \
|
|
the default location (``libbmc/external/cermine.jar``).
|
|
:returns: A dict of cleaned plaintext citations and their associated DOI.
|
|
"""
|
|
# TODO:
|
|
# * Do not convert to plain text, but use the extra metadata from
|
|
# CERMINE
|
|
# Call CERMINE on the PDF file
|
|
cermine_output = cermine(pdf_file, force_API, override_local)
|
|
# Parse the resulting XML
|
|
root = ET.fromstring(cermine_output)
|
|
plaintext_references = [
|
|
# Remove extra whitespaces
|
|
tools.clean_whitespaces(
|
|
# Convert XML element to string, discarding any leading "[n]"
|
|
ET.tostring(e, method="text").decode("utf-8").replace(e.text, ""))
|
|
for e in root.iter("mixed-citation")]
|
|
# Call the plaintext methods to fetch DOIs
|
|
return plaintext.get_cited_DOIs(plaintext_references)
|
|
|
|
|
|
def grobid(pdf_file):
|
|
"""
|
|
Run `Grobid <https://github.com/kermitt2/grobid>`_ on a given PDF file to \
|
|
extract references.
|
|
|
|
.. note::
|
|
|
|
Before using this function, you have to download and build Grobid on \
|
|
your system. See \
|
|
`<https://grobid.readthedocs.org/en/latest/Install-Grobid/>`_ \
|
|
for more infos on this. You need Java and \
|
|
``grobid-core-`<current version>`.one-jar.jar`` to be in your \
|
|
``$PATH``.
|
|
|
|
:param pdf_file: Path to the PDF file to handle.
|
|
:returns: Raw output from ``Grobid`` or ``None`` if an error occurred.
|
|
"""
|
|
# TODO + update docstring
|
|
# TODO: Use https://github.com/kermitt2/grobid-example
|
|
subprocess.check_output(["java",
|
|
"-jar", "grobid-core-0.3.0.one-jar.jar",
|
|
"-Xmx1024m", # Avoid OutOfMemoryException
|
|
"-gH", "/path/to/Grobid/grobid/grobid-home",
|
|
"-gP", "/path/to/Grobid/grobid-home/config/grobid.properties",
|
|
"-dIn", "/path/to/input/directory",
|
|
"-dOut", "/path/to/output/directory",
|
|
"-exe", "processReferences"])
|
|
|
|
|
|
def pdfextract(pdf_file):
|
|
"""
|
|
Run `pdfextract <https://github.com/CrossRef/pdfextract>`_ on a given PDF \
|
|
file to extract references.
|
|
|
|
.. note::
|
|
|
|
Before using this function, you have to install pdfextract on \
|
|
your system. See \
|
|
`<https://github.com/CrossRef/pdfextract#quick-start>`_ \
|
|
for more infos on this. You need the ``pdf-extract`` command \
|
|
to be in your ``$PATH``. This can be done easily using \
|
|
``gem install pdf-extract``, provided that you have a correct \
|
|
Ruby install on your system.
|
|
|
|
.. note::
|
|
|
|
``pdfextract`` is full a bugs and as the time of writing this, \
|
|
you had to manually ``gem install pdf-reader -v 1.2.0`` \
|
|
before installing ``pdfextract`` or you would get errors. See \
|
|
`this Github issue <https://github.com/CrossRef/pdfextract/issues/23>`_.
|
|
|
|
:param pdf_file: Path to the PDF file to handle.
|
|
:returns: Raw output from ``pdfextract`` or ``None`` if an error \
|
|
occurred. No post-processing is done. See \
|
|
``libbmc.citations.pdf.pdfextract_dois`` for a similar function \
|
|
with post-processing to return DOIs.
|
|
"""
|
|
try:
|
|
# Run pdf-extract
|
|
references = subprocess.check_output(["pdf-extract",
|
|
"extract", "--references",
|
|
pdf_file])
|
|
return references
|
|
except subprocess.CalledProcessError:
|
|
return None
|
|
|
|
|
|
def pdfextract_dois(pdf_file):
|
|
"""
|
|
Extract DOIs of references using \
|
|
`pdfextract <https://github.com/CrossRef/pdfextract>`_.
|
|
|
|
.. note::
|
|
|
|
See ``libbmc.citations.pdf.pdfextract`` function as this one is just \
|
|
a wrapper around it.
|
|
See ``libbmc.citations.plaintext.get_cited_dois`` as well for the \
|
|
returned value, as it is ultimately called by this function.
|
|
|
|
:param pdf_file: Path to the PDF file to handle.
|
|
:returns: A dict of cleaned plaintext citations and their associated DOI.
|
|
"""
|
|
# Call pdf-extract on the PDF file
|
|
references = pdfextract(pdf_file)
|
|
# Parse the resulting XML
|
|
root = ET.fromstring(references)
|
|
plaintext_references = [e.text for e in root.iter("reference")]
|
|
# Call the plaintext methods to fetch DOIs
|
|
return plaintext.get_cited_DOIs(plaintext_references)
|