class TEIExtractor:
def __init__(self, file, db, xml, data_file, test_tsv=None):
self.file = file
self.xml = xml
self.uni_rank = ReadPickle('uni_rank.pickle')
self.sjr = ReadPickle('journal_dictionary.pkl')
self.document = test_tsv
self.test_csv = data_file
self.paper = Paper()
with open(file, 'rb') as tei:
self.soup = BeautifulSoup(tei, features="lxml")
self.db=db
# TODO: return paper | redesign extractor to make it more modular to test individual components
def get_self_citations(self):
# DOI
doi = self.soup.teiheader.find("idno", type="DOI")
if doi:
self.paper.doi = elem_to_text(doi)
elif self.document:
self.paper.doi = self.document['doi']
# Title
title = self.soup.teiheader.find("title")
if title:
self.paper.title = elem_to_text(title)
# Authors
authors = self.get_authors(self.soup.analytic.find_all('author'))
if authors:
self.paper.authors = authors
if self.soup.abstract:
self.paper.abstract = elem_to_text(self.soup.abstract)
# Citations
bibliography = self.soup.listbibl.find_all('biblstruct')
for bibl in bibliography:
citation = Citation()
cited_paper = bibl.analytic
if cited_paper:
citation.title = elem_to_text(cited_paper.find("title", type="main"))
citation_authors = self.get_authors(cited_paper.find_all("author"))
citation.doi = elem_to_text(cited_paper.find("idno", type="DOI"))
if citation_authors:
citation.authors = citation_authors
cited_journal = bibl.monogr
if cited_journal:
citation.source = elem_to_text(cited_journal.find("title"))
try:
citation.publish_year = cited_journal.imprint.date['when']
except TypeError:
pass
self.paper.citations.append(citation)
self.paper.set_self_citations()
return {'doi': self.paper.doi, 'title': self.paper.title, 'total_citations': len(self.paper.citations),
'self_citations': self.paper.self_citations, 'abstract': self.paper.abstract}
def get_reading_score(self, abstract):
if isinstance(abstract,str):
if not abstract: return 0
return textstat.flesch_reading_ease(abstract)
return 0
def get_subjectivity(self, abstract):
if isinstance(abstract,str):
if len(abstract)<10: return -1
txtblob = TextBlob(abstract)
return txtblob.sentiment.subjectivity
return -1
def get_sentiment(self, abstract):
if isinstance(abstract,str):
if len(abstract)<10: return -1
label = predictor.predict(abstract)['label']
return int(label)
return -1
def extract_paper_info(self):
# DOI
doi = self.soup.teiheader.find("idno", type="DOI")
if doi:
self.paper.doi = elem_to_text(doi)
elif self.document:
self.paper.doi = self.document['doi']
# Title
title = self.soup.teiheader.find("title")
if title:
self.paper.title = elem_to_text(title)
# Authors
authors = self.get_authors(self.soup.analytic.find_all('author'))
if authors:
self.paper.authors = authors
if self.soup.abstract:
self.paper.abstract = elem_to_text(self.soup.abstract)
# Year
published = self.soup.analytic.find("publicationstmt")
if published:
self.paper.year = elem_to_text(published.find("date", type="when"))
# Organization / Affiliations
affiliations = self.soup.analytic.find_all('affiliation')
for affiliation in affiliations:
org = Organization()
org.type = "institution"
org.name = elem_to_text(affiliation.find("orgname", type="institution"))
address = Address()
addr = affiliation.find("address")
if addr:
address.place = elem_to_text(addr.find("settlement"))
address.region = elem_to_text(addr.find("region"))
address.country = elem_to_text(addr.find("country"))
org.address = address
self.paper.affiliations.append(org)
# University Ranking
if self.paper.affiliations:
if self.paper.affiliations[0] != '':
self.paper.uni_rank = self.uni_rank.get_rank(self.paper.affiliations[0].name)
elif len(self.paper.affiliations) > 1:
self.paper.uni_rank = self.uni_rank.get_rank(self.paper.affiliations[1].name)
else:
self.paper.uni_rank = self.uni_rank.get_rank('Random')
# Citations
bibliography = self.soup.listbibl.find_all('biblstruct')
for bibl in bibliography:
citation = Citation()
cited_paper = bibl.analytic
if cited_paper:
citation.title = elem_to_text(cited_paper.find("title", type="main"))
citation_authors = self.get_authors(cited_paper.find_all("author"))
citation.doi = elem_to_text(cited_paper.find("idno", type="DOI"))
if citation_authors:
citation.authors = citation_authors
cited_journal = bibl.monogr
if cited_journal:
citation.source = elem_to_text(cited_journal.find("title"))
try:
citation.publish_year = cited_journal.imprint.date['when']
except TypeError:
pass
self.paper.citations.append(citation)
# NER - Ack pairs - Funding status
self.paper.ack_pairs = self.get_funding_status()
er_list = [org for (entity, org) in self.paper.ack_pairs]
if 'ORG' in er_list:
self.paper.funded = 1
else:
self.paper.funded = 0
# SJR
api_resp = self.get_sjr(self.paper.doi, self.paper.title,self.db)
# Adding the paragraphs from the paper to the corpus
extractor = ClaimEvidenceExtractor(self.xml, self.soup,self.test_csv)
os.chdir("/scifact/")
extractor.make_corpus()
# Get response for claim evidence using request to API
response = requests.get('http://0.0.0.0:8000/getclaimevidence')
print(response)
self.support, self.refute, self.ratio = extractor.get_results()
print('support:',self.support,'refute:',self.refute,'ratio:',self.ratio)
os.chdir("../")
if api_resp:
self.paper.cited_by_count = api_resp["num_citations"]
self.paper.sjr = api_resp["sjr"]
self.paper.subject = api_resp["subject"]
self.paper.subject_code = api_resp["subject_code"]
self.paper.normalized = api_resp["normalized_citations"]
self.paper.velocity = api_resp["citationVelocity"]
self.paper.influentialcitations = api_resp["influentialCitationCount"]
self.paper.references = api_resp["references_count"]
self.paper.flag = api_resp["openaccessflag"]
self.paper.influentialref = api_resp["influentialReferencesCount"]
self.paper.ref_background = api_resp["reference_background"]
self.paper.ref_result = api_resp["reference_result"]
self.paper.ref_method = api_resp["reference_methodology"]
self.paper.cite_background = api_resp["citations_background"]
self.paper.cite_result = api_resp["citations_result"]
self.paper.cite_method = api_resp["citations_methodology"]
self.paper.cite_next = api_resp["citations_next"]
self.paper.influential_references_methodology = api_resp["upstream_influential_methodology_count"]
self.paper.issn = api_resp["ISSN"]
self.paper.auth = api_resp["authors"]
self.paper.age = api_resp["age"]
if api_resp["abstract"]:
self.paper.abstract = api_resp["abstract"]
# Set self-citations
self.paper.self_citations = self.paper.set_self_citations()
# return paper
#calculate coCitations
t2,t3 = coCite(self.paper.doi, self.db)
#calculate NLP features
reading_score = self.get_reading_score(self.paper.abstract)
subjectivity = self.get_subjectivity(self.paper.abstract)
sentiment = self.get_sentiment(self.paper.abstract)
return {"doi": self.paper.doi, "title": self.paper.title, "num_citations": self.paper.cited_by_count,
"author_count": len(self.paper.authors),"sjr": self.paper.sjr, "u_rank": self.paper.uni_rank,
"funded": self.paper.funded,"self_citations": self.paper.self_citations, "subject": self.paper.subject,
"subject_code": self.paper.subject_code, "citationVelocity": self.paper.velocity,
"influentialCitationCount": self.paper.influentialcitations, "references_count": self.paper.references,
"openaccessflag": self.paper.flag, "influentialReferencesCount": self.paper.influentialref,
"normalized_citations": self.paper.normalized, "reference_background": self.paper.ref_background,
"reference_result": self.paper.ref_result, "reference_methodology": self.paper.ref_method,
"citations_background": self.paper.cite_background, "citations_result": self.paper.cite_result,
"citations_methodology": self.paper.cite_method, "citations_next": self.paper.cite_next,
"upstream_influential_methodology_count": self.paper.influential_references_methodology,
"coCite2":t2, "coCite3":t3, "ISSN":self.paper.issn, "authors":self.paper.auth,"citations":api_resp["citations"],"age":self.paper.age,
"reading_score":reading_score, "subjectivity":subjectivity, "sentiment":sentiment, "supporting_sentences":self.support, "refuting_sentences":self.refute, "ratio_support":self.ratio}
@staticmethod
def get_authors(authors):
authors_list = []
for author in authors:
person = Author()
pers_name = author.persname
if not pers_name:
continue
person.first_name = elem_to_text(pers_name.find("forename", type="first"))
person.middle_name = elem_to_text(pers_name.find("forename", type="middle"))
person.surname = elem_to_text(pers_name.surname)
person.set_name()
if not any(auth.name == person.name for auth in authors_list):
authors_list.append(person)
return authors_list
def get_funding_status(self):
pairs = NER(XML2ack(self.file))
return pairs
@staticmethod
def get_sjr(doi,title,db):
response = getsemantic(doi,title,db)
crossref = response.get_row()
scopus_search = response.return_search()
serial_title = response.return_serialtitle()
semantic = response.return_semantic()
#pdb.set_trace()
final = {"doi":response.doi, "title":response.title, "sjr": response.sjr, "num_citations": response.citedby,"subject":response.subject,"subject_code":response.subject_code,"normalized_citations":response.normalized,"citationVelocity":response.velocity,"influentialCitationCount":response.incite,"references_count":response.refcount,"openaccessflag":response.openaccess,"influentialReferencesCount":response.inref, "reference_background": response.refback, "reference_result":response.refresult, "reference_methodology":response.refmeth,"citations_background":response.cback,"citations_result":response.cresult,"citations_methodology":response.cmeth, "citations_next":response.next, "upstream_influential_methodology_count": response.upstream_influential_methodology_count, "ISSN": response.issn, "authors":response.auth, "citations":response.citations,"age":response.age,"abstract":response.ab}
return final
class TEIExtractor:
def __init__(self, file, test_tsv=None):
self.file = file
self.uni_rank = ReadPickle('uni_rank.pickle')
self.sjr = ReadPickle('journal_dictionary.pkl')
self.document = test_tsv
self.paper = Paper()
with open(file, 'rb') as tei:
self.soup = BeautifulSoup(tei, features="lxml")
# TODO: return paper | redesign extractor to make it more modular to test individual components
def get_self_citations(self):
# DOI
doi = self.soup.teiheader.find("idno", type="DOI")
if doi:
self.paper.doi = elem_to_text(doi)
elif self.document:
self.paper.doi = self.document['doi']
# Title
title = self.soup.teiheader.find("title")
if title:
self.paper.title = elem_to_text(title)
# Authors
authors = self.get_authors(self.soup.analytic.find_all('author'))
if authors:
self.paper.authors = authors
# Citations
bibliography = self.soup.listbibl.find_all('biblstruct')
for bibl in bibliography:
citation = Citation()
cited_paper = bibl.analytic
if cited_paper:
citation.title = elem_to_text(
cited_paper.find("title", type="main"))
citation_authors = self.get_authors(
cited_paper.find_all("author"))
citation.doi = elem_to_text(
cited_paper.find("idno", type="DOI"))
if citation_authors:
citation.authors = citation_authors
cited_journal = bibl.monogr
if cited_journal:
citation.source = elem_to_text(cited_journal.find("title"))
try:
citation.publish_year = cited_journal.imprint.date['when']
except TypeError:
pass
self.paper.citations.append(citation)
self.paper.set_self_citations()
return {
'doi': self.paper.doi,
'title': self.paper.title,
'total_citations': len(self.paper.citations),
'self_citations': self.paper.self_citations
}
def extract_paper_info(self):
# DOI
doi = self.soup.teiheader.find("idno", type="DOI")
if doi:
self.paper.doi = elem_to_text(doi)
elif self.document:
self.paper.doi = self.document['doi']
# Title
title = self.soup.teiheader.find("title")
if title:
self.paper.title = elem_to_text(title)
# Authors
authors = self.get_authors(self.soup.analytic.find_all('author'))
if authors:
self.paper.authors = authors
# Year
published = self.soup.analytic.find("publicationstmt")
if published:
self.paper.year = elem_to_text(published.find("date", type="when"))
# Organization / Affiliations
affiliations = self.soup.analytic.find_all('affiliation')
for affiliation in affiliations:
org = Organization()
org.type = "institution"
org.name = elem_to_text(
affiliation.find("orgname", type="institution"))
address = Address()
addr = affiliation.find("address")
if addr:
address.place = elem_to_text(addr.find("settlement"))
address.region = elem_to_text(addr.find("region"))
address.country = elem_to_text(addr.find("country"))
org.address = address
self.paper.affiliations.append(org)
# University Ranking
if self.paper.affiliations:
if self.paper.affiliations[0] != '':
self.paper.uni_rank = self.uni_rank.get_rank(
self.paper.affiliations[0].name)
elif len(self.paper.affiliations) > 1:
self.paper.uni_rank = self.uni_rank.get_rank(
self.paper.affiliations[1].name)
else:
self.paper.uni_rank = self.uni_rank.get_rank('Random')
# Citations
bibliography = self.soup.listbibl.find_all('biblstruct')
for bibl in bibliography:
citation = Citation()
cited_paper = bibl.analytic
if cited_paper:
citation.title = elem_to_text(
cited_paper.find("title", type="main"))
citation_authors = self.get_authors(
cited_paper.find_all("author"))
citation.doi = elem_to_text(
cited_paper.find("idno", type="DOI"))
if citation_authors:
citation.authors = citation_authors
cited_journal = bibl.monogr
if cited_journal:
citation.source = elem_to_text(cited_journal.find("title"))
try:
citation.publish_year = cited_journal.imprint.date['when']
except TypeError:
pass
self.paper.citations.append(citation)
# NER - Ack pairs - Funding status
self.paper.ack_pairs = self.get_funding_status()
er_list = [org for (entity, org) in self.paper.ack_pairs]
if 'ORG' in er_list:
self.paper.funded = 1
else:
self.paper.funded = 0
# SJR
api_resp = self.get_sjr(self.paper.doi, self.paper.title)
if api_resp:
self.paper.cited_by_count = api_resp["num_citations"]
self.paper.sjr = api_resp["sjr"]
self.paper.subject = api_resp["subject"]
self.paper.subject_code = api_resp["subject_code"]
self.paper.normalized = api_resp["normalized_citations"]
self.paper.velocity = api_resp["citationVelocity"]
self.paper.influentialcitations = api_resp[
"influentialCitationCount"]
self.paper.references = api_resp["references_count"]
self.paper.flag = api_resp["openaccessflag"]
self.paper.influentialref = api_resp["influentialReferencesCount"]
self.paper.ref_background = api_resp["reference_background"]
self.paper.ref_result = api_resp["reference_result"]
self.paper.ref_method = api_resp["reference_methodology"]
self.paper.cite_background = api_resp["citations_background"]
self.paper.cite_result = api_resp["citations_result"]
self.paper.cite_method = api_resp["citations_methodology"]
self.paper.cite_next = api_resp["citations_next"]
# Set self-citations
self.paper.self_citations = self.paper.set_self_citations()
# Set influential_methodology_references
self.paper.influential_references_methodology = self.set_influential_references_methodology(
)
# return paper
t2, t3 = coCite(self.paper.doi)
return {
"doi": self.paper.doi,
"title": self.paper.title,
"num_citations": self.paper.cited_by_count,
"author_count": len(self.paper.authors),
"sjr": self.paper.sjr,
"u_rank": self.paper.uni_rank,
"funded": self.paper.funded,
"self_citations": self.paper.self_citations,
"subject": self.paper.subject,
"subject_code": self.paper.subject_code,
"citationVelocity": self.paper.velocity,
"influentialCitationCount": self.paper.influentialcitations,
"references_count": self.paper.references,
"openaccessflag": self.paper.flag,
"influentialReferencesCount": self.paper.influentialref,
"normalized_citations": self.paper.normalized,
"reference_background": self.paper.ref_background,
"reference_result": self.paper.ref_result,
"reference_methodology": self.paper.ref_method,
"citations_background": self.paper.cite_background,
"citations_result": self.paper.cite_result,
"citations_methodology": self.paper.cite_method,
"citations_next": self.paper.cite_next,
"upstream_influential_methodology_count":
self.paper.influential_references_methodology,
"coCite2": t2,
"coCite3": t3
}
@staticmethod
def get_authors(authors):
authors_list = []
for author in authors:
person = Author()
pers_name = author.persname
if not pers_name:
continue
person.first_name = elem_to_text(
pers_name.find("forename", type="first"))
person.middle_name = elem_to_text(
pers_name.find("forename", type="middle"))
person.surname = elem_to_text(pers_name.surname)
person.set_name()
if not any(auth.name == person.name for auth in authors_list):
authors_list.append(person)
return authors_list
def get_funding_status(self):
pairs = NER(XML2ack(self.file))
return pairs
@staticmethod
def get_sjr(doi, title):
api = getapi(doi, title)
if api.empty:
return None
else:
try:
cited_by = api['num_citations'][0]
except KeyError:
cited_by = 0
try:
normalized = api['normalized_citations'][0]
except:
normalized = 0.0
try:
velocity = api['citationVelocity'][0]
except:
velocity = 0
try:
influentialcitations = api['influentialCitationCount'][0]
except:
influentialcitations = 0
try:
references = api['references_count'][0]
except:
references = 0
try:
sjr_score = api['SJR'][0]
except KeyError:
sjr_score = 0
try:
subject = api['subject'][0]
except:
subject = 0
try:
subject_code = api['subject_code'][0]
except:
subject_code = 900
try:
flag = api['openaccessflag'][0]
except:
flag = 0
try:
influentialref = api['influentialReferencesCount'][0]
except:
influentialref = 0
try:
ref_background = api['reference_background'][0]
except:
ref_background = 0
try:
ref_result = api['reference_result'][0]
except:
ref_result = 0
try:
ref_method = api['reference_methodology'][0]
except:
ref_method = 0
try:
cite_background = api['citations_background'][0]
except:
cite_background = 0
try:
cite_result = api['citations_result'][0]
except:
cite_result = 0
try:
cite_method = api['citations_methodology'][0]
except:
cite_method = 0
try:
cite_next = api['citation_next'][0]
except:
cite_next = 0
return {
"sjr": sjr_score,
"num_citations": cited_by,
"subject": subject,
"subject_code": subject_code,
"normalized_citations": normalized,
"citationVelocity": velocity,
"influentialCitationCount": influentialcitations,
"references_count": references,
"openaccessflag": flag,
"influentialReferencesCount": influentialref,
"reference_background": ref_background,
"reference_result": ref_result,
"reference_methodology": ref_method,
"citations_background": cite_background,
"citations_result": cite_result,
"citations_methodology": cite_method,
"citations_next": cite_next
}
def set_influential_references_methodology(self):
# Counts the number of influential references in the paper in the context of methodology
count = 0
if self.paper.doi:
url = 'https://partner.semanticscholar.org/v1/paper/{0}'.format(
self.paper.doi)
headers = {'x-api-key': 'I6SO5Ckndk67RitJNJOFR4d7jDiVpWOgaMFUhgkM'}
response_payload = requests.get(url, headers=headers).json()
try:
references = response_payload['references']
for reference in references:
try:
if 'methodology' in reference['intent'] and reference[
'isInfluential']:
count += 1
except KeyError:
continue
except KeyError:
pass
return count
