aboutsummaryrefslogtreecommitdiffstats
path: root/python/parse_pubmed_xml.py
diff options
context:
space:
mode:
Diffstat (limited to 'python/parse_pubmed_xml.py')
-rw-r--r--python/parse_pubmed_xml.py370
1 files changed, 370 insertions, 0 deletions
diff --git a/python/parse_pubmed_xml.py b/python/parse_pubmed_xml.py
new file mode 100644
index 00000000..9350e9a4
--- /dev/null
+++ b/python/parse_pubmed_xml.py
@@ -0,0 +1,370 @@
+
+import sys
+import json
+import datetime
+from bs4 import BeautifulSoup
+from bs4.element import NavigableString
+
+# from: https://www.ncbi.nlm.nih.gov/books/NBK3827/table/pubmedhelp.T.publication_types/?report=objectonly
+PUBMED_RELEASE_TYPE_MAP = {
+ #Adaptive Clinical Trial
+ "Address": "speech",
+ "Autobiography": "book",
+ #Bibliography
+ "Biography": "book",
+ #Case Reports
+ "Classical Article": "article-journal",
+ #Clinical Conference
+ #Clinical Study
+ #Clinical Trial
+ #Clinical Trial, Phase I
+ #Clinical Trial, Phase II
+ #Clinical Trial, Phase III
+ #Clinical Trial, Phase IV
+ #Clinical Trial Protocol
+ #Clinical Trial, Veterinary
+ #Collected Works
+ #Comparative Study
+ #Congress
+ #Consensus Development Conference
+ #Consensus Development Conference, NIH
+ #Controlled Clinical Trial
+ "Dataset": "dataset",
+ #Dictionary
+ #Directory
+ #Duplicate Publication
+ "Editorial": "editorial",
+ #English Abstract # doesn't indicate that this is abstract-only
+ #Equivalence Trial
+ #Evaluation Studies
+ #Expression of Concern
+ #Festschrift
+ #Government Document
+ #Guideline
+ "Historical Article": "article-journal",
+ #Interactive Tutorial
+ "Interview": "interview",
+ "Introductory Journal Article": "article-journal",
+ "Journal Article": "article-journal",
+ "Lecture": "speech",
+ "Legal Case": "legal_case",
+ "Legislation": "legislation",
+ "Letter": "letter",
+ #Meta-Analysis
+ #Multicenter Study
+ #News
+ "Newspaper Article": "article-newspaper",
+ #Observational Study
+ #Observational Study, Veterinary
+ #Overall
+ #Patient Education Handout
+ #Periodical Index
+ #Personal Narrative
+ #Portrait
+ #Practice Guideline
+ #Pragmatic Clinical Trial
+ #Publication Components
+ #Publication Formats
+ #Publication Type Category
+ #Randomized Controlled Trial
+ #Research Support, American Recovery and Reinvestment Act
+ #Research Support, N.I.H., Extramural
+ #Research Support, N.I.H., Intramural
+ #Research Support, Non-U.S. Gov't Research Support, U.S. Gov't, Non-P.H.S.
+ #Research Support, U.S. Gov't, P.H.S.
+ #Review # in the "literature review" sense, not "product review"
+ #Scientific Integrity Review
+ #Study Characteristics
+ #Support of Research
+ #Systematic Review
+ "Technical Report": "report",
+ #Twin Study
+ #Validation Studies
+ #Video-Audio Media
+ #Webcasts
+}
+
+MONTH_ABBR_MAP = {
+ "Jan": 1, "01": 1,
+ "Feb": 2, "02": 2,
+ "Mar": 3, "03": 3,
+ "Apr": 4, "04": 4,
+ "May": 5, "05": 5,
+ "Jun": 6, "06": 6,
+ "Jul": 7, "07": 7,
+ "Aug": 8, "08": 8,
+ "Sep": 9, "09": 9,
+ "Oct": 10, "10": 10,
+ "Nov": 11, "11": 11,
+ "Dec": 12, "12": 12,
+}
+
+class PubMedParser():
+ """
+ Converts PubMed/MEDLINE XML into in release entity (which can dump as JSON)
+
+ TODO: MEDLINE doesn't include PMC/OA license; could include in importer?
+ TODO: clean (ftfy) title, original title, etc
+ """
+
+ def __init__(self):
+ pass
+
+ def parse_file(self, handle):
+
+ # 1. open with beautiful soup
+ soup = BeautifulSoup(handle, "xml")
+
+ # 2. iterate over articles, call parse_article on each
+ for article in soup.find_all("PubmedArticle"):
+ resp = self.parse_article(article)
+ print(json.dumps(resp))
+ #sys.exit(-1)
+
+ def parse_article(self, a):
+
+ medline = a.MedlineCitation
+ # PubmedData isn't required by DTD, but seems to always be present
+ pubmed = a.PubmedData
+ extra = dict()
+ extra_pubmed = dict()
+
+ identifiers = pubmed.ArticleIdList
+ doi = identifiers.find("ArticleId", IdType="doi")
+ if doi:
+ doi = doi.string.lower()
+
+ pmcid = identifiers.find("ArticleId", IdType="pmc")
+ if pmcid:
+ pmcid = pmcid.string
+
+ release_type = None
+ for pub_type in medline.Article.PublicationTypeList.find_all("PublicationType"):
+ if pub_type.string in PUBMED_RELEASE_TYPE_MAP:
+ release_type = PUBMED_RELEASE_TYPE_MAP[pub_type.string]
+ break
+ if medline.Article.PublicationTypeList.find(string="Retraction of Publication"):
+ release_type = "retraction"
+ retraction_of = medline.find("CommentsCorrections", RefType="RetractionOf")
+ if retraction_of:
+ extra_pubmed['retraction_of_raw'] = retraction_of.RefSource.string
+ extra_pubmed['retraction_of_pmid'] = retraction_of.PMID.string
+
+ # everything in medline is published
+ release_status = "published"
+ if medline.Article.PublicationTypeList.find(string="Corrected and Republished Article"):
+ release_status = "updated"
+ if medline.Article.PublicationTypeList.find(string="Retracted Publication"):
+ release_status = "retracted"
+
+ pages = medline.find('MedlinePgn')
+ if pages:
+ pages = pages.string
+
+ title = medline.Article.ArticleTitle.string, # always present
+ if type(title) is tuple:
+ title = ': '.join(title)
+ if title.endswith('.'):
+ title = title[:-1]
+ # this hides some "special" titles, but the vast majority are
+ # translations; translations don't always include the original_title
+ if title.startswith('[') and title.endswith(']'):
+ title = title[1:-1]
+
+ original_title = medline.Article.find("VernacularTitle", recurse=False)
+ if original_title:
+ original_title = original_title.string
+ if original_title.endswith('.'):
+ original_title = original_title[:-1]
+
+ # TODO: happening in alpha order, not handling multi-language well.
+ # also need to convert lang codes: https://www.nlm.nih.gov/bsd/language_table.html
+ language = medline.Article.Language
+ if language:
+ language = language.string
+ # TODO: map to two-letter
+ if language in ("und", "un"):
+ # "undetermined"
+ language = None
+
+ ### Journal/Issue Metadata
+ # MedlineJournalInfo is always present
+ container = dict()
+ container_extra = dict()
+ mji = medline.MedlineJournalInfo
+ if mji.find("Country"):
+ container_extra['country_name'] = mji.Country.string
+ if mji.find("ISSNLinking"):
+ container['issnl'] = mji.ISSNLinking.string
+
+ journal = medline.Article.Journal
+ issnp = journal.find("ISSN", IssnType="Print")
+ if issnp:
+ container_extra['issnp'] = issnp.string
+
+ pub_date = journal.PubDate
+ release_date = None
+ if pub_date.find("MedlineDate"):
+ release_year = int(pub_date.MedlineDate.string.split()[0][:4])
+ else:
+ release_year = int(pub_date.Year.string)
+ if pub_date.find("Day") and pub_date.find("Month"):
+ release_date = datetime.date(
+ release_year,
+ MONTH_ABBR_MAP[pub_date.Month.string],
+ int(pub_date.Day.string))
+ release_date = release_date.isoformat()
+
+ ji = journal.JournalIssue
+ volume = None
+ if ji.find("Volume"):
+ volume = ji.Volume.string
+ issue = None
+ if ji.find("Issue"):
+ issue = ji.Issue.string
+ if journal.find("Title"):
+ container['name'] = journal.Title.string
+
+ if extra_pubmed:
+ extra['pubmed'] = extra_pubmed
+ if not extra:
+ extra = None
+
+ ### Abstracts
+ # "All abstracts are in English"
+ abstracts = []
+ first_abstract = medline.find("AbstractText")
+ if first_abstract and first_abstract.get('NlmCategory'):
+ joined = "\n".join([m.get_text() for m in medline.find_all("AbstractText")])
+ abstracts.append(dict(
+ content=joined,
+ mimetype="text/plain",
+ lang="en",
+ ))
+ else:
+ for abstract in medline.find_all("AbstractText"):
+ abstracts.append(dict(
+ content=abstract.get_text().strip(),
+ mimetype="text/plain",
+ lang="en",
+ ))
+ if abstract.find('math'):
+ abstracts.append(dict(
+ # strip the <AbstractText> tags
+ content=str(abstract)[14:-15],
+ mimetype="application/mathml+xml",
+ lang="en",
+ ))
+ if not abstracts:
+ abstracts = None
+
+ ### Contribs
+ contribs = []
+ if medline.AuthorList:
+ for author in medline.AuthorList.find_all("Author"):
+ contrib = dict(
+ role="author",
+ )
+ if author.ForeName:
+ contrib['raw_name'] = "{} {}".format(author.ForeName.string, author.LastName.string)
+ elif author.LastName:
+ contrib['raw_name'] = author.LastName.string
+ contrib_extra = dict()
+ orcid = author.find("Identifier", Source="ORCID")
+ if orcid:
+ # needs re-formatting from, eg, "0000000179841889"
+ orcid = orcid.string
+ if orcid.startswith("http://orcid.org/"):
+ orcid = orcid.replace("http://orcid.org/", "")
+ elif orcid.startswith("https://orcid.org/"):
+ orcid = orcid.replace("https://orcid.org/", "")
+ elif not '-' in orcid:
+ orcid = "{}-{}-{}-{}".format(
+ orcid[0:4],
+ orcid[4:8],
+ orcid[8:12],
+ orcid[12:16],
+ )
+ contrib_extra['orcid'] = orcid
+ affiliation = author.find("Affiliation")
+ if affiliation:
+ contrib['raw_affiliation'] = affiliation.string
+ if author.find("EqualContrib"):
+ # TODO: schema for this?
+ contrib_extra['equal_contrib'] = True
+ if contrib_extra:
+ contrib['extra'] = contrib_extra
+ contribs.append(contrib)
+
+ if medline.AuthorList['CompleteYN'] == 'N':
+ contribs.append(dict(raw_name="et al."))
+ if not contribs:
+ contribs = None
+
+ ### References
+ refs = []
+ if pubmed.ReferenceList:
+ for ref in pubmed.ReferenceList.find_all('Reference'):
+ ref_obj = dict()
+ ref_extra = dict()
+ ref_pmid = ref.find("ArticleId", IdType="pubmed")
+ if ref_pmid:
+ ref_extra['pmid'] = ref_pmid.string
+ ref_raw = ref.Citation
+ if ref_raw:
+ ref_extra['raw'] = ref_raw.string
+ if ref_extra:
+ ref_obj['extra'] = ref_extra
+ refs.append(ref_obj)
+ if not refs:
+ refs = None
+
+ re = dict(
+ work_id=None,
+ title=title,
+ original_title=original_title,
+ release_type=release_type,
+ release_status=release_status,
+ release_date=release_date,
+ release_year=release_year,
+ doi=doi,
+ pmid=int(medline.PMID.string), # always present
+ pmcid=pmcid,
+ #isbn13 # never in Article
+ volume=volume,
+ issue=issue,
+ pages=pages,
+ #publisher # not included?
+ language=language,
+ #license_slug # not in MEDLINE
+
+ # content, mimetype, lang
+ abstracts=abstracts,
+
+ # raw_name, role, raw_affiliation, extra
+ contribs=contribs,
+
+ # key, year, container_name, title, locator
+ # extra: volume, authors, issue, publisher, identifiers
+ refs=refs,
+
+ # name, type, publisher, issnl
+ # extra: issnp, issne, original_name, languages, country
+ container=container,
+
+ # extra:
+ # withdrawn_date
+ # translation_of
+ # subtitle
+ # aliases
+ # container_name
+ # group-title
+ # pubmed: retraction refs
+ extra=extra,
+ )
+
+ return re
+
+if __name__=='__main__':
+ parser = PubMedParser()
+ parser.parse_file(open(sys.argv[1]))