def loadPaperSentence(self, s, newDocument, parent):
"""
Given a string, adds the sentence to the SciDoc, parses the citations,
matches them with the references
Args:
s: string
newDocument: SciDoc
parent: id of element this sentence will hang from (p)
"""
def replaceTempCitToken(s, temp, final):
"""
replace temporary citation placeholder with permanent one
"""
return re.sub(CITATION_FORM % temp, CITATION_FORM % final, annotated_s, flags=re.IGNORECASE)
newSent=newDocument.addSentence(parent,"")
annotated_s,citations_found=annotateCitationsInSentence(s, newDocument.metadata["original_citation_style"])
annotated_citations=[]
if newDocument.metadata["original_citation_style"]=="APA":
for index,citation in enumerate(citations_found):
newCit=newDocument.addCitation(sent_id=newSent["id"])
reference=matchCitationWithReference(citation, newDocument["references"])
## print (citation["text"]," -> ", formatReference(reference))
if reference:
newCit["ref_id"]=reference["id"]
else:
# do something else?
newCit["ref_id"]=None
annotated_citations.append(newCit)
annotated_s=replaceTempCitToken(annotated_s, index+1, newCit["id"])
elif newDocument.metadata["original_citation_style"]=="AFI":
for index,citation in enumerate(citations_found):
newCit=newDocument.addCitation(sent_id=newSent["id"])
# TODO check this: maybe not this simple. May need matching function.
newCit["ref_id"]="ref"+str(int(citation["num"])-1)
annotated_citations.append(newCit)
annotated_s=replaceTempCitToken(annotated_s, index+1, newCit["id"])
newSent["citations"]=[acit["id"] for acit in annotated_citations]
newSent["text"]=annotated_s
# deal with many citations within characters of each other: make them know they are a cluster
# TODO cluster citations? Store them in some other way?
newDocument.countMultiCitations(newSent)
def annotatePlainTextCitations(self, s, newDocument, newSent):
"""
If the citations aren't tagged with <xref> because Sapienta stripped
them away, try to extract the citations from plain text. *sigh*
:param s: BeautifulSoup tag of the sentence
:param newDocument: SciDoc instance we are populating
:param newSent: new sentence in this document that we are adding
"""
def replaceTempCitToken(s, temp, final):
"""
replace temporary citation placeholder with permanent one
"""
return re.sub(CITATION_FORM % temp, CITATION_FORM % final, annotated_s, flags=re.IGNORECASE)
def replaceTempCitTokenMulti(s, temp, final_list):
"""
Replace temporary citation placeholder with a list of permanent
ones to deal with multi citations, e.g. [1,2,3]
"""
assert(isinstance(final_list, list))
rep_string="".join([CITATION_FORM % final for final in final_list])
return re.sub(CITATION_FORM % temp, rep_string, annotated_s, flags=re.IGNORECASE)
if not newDocument.metadata.get("original_citation_style", None):
newDocument.metadata["original_citation_style"]="AFI"
annotated_s,citations_found=annotateCitationsInSentence(s, newDocument.metadata["original_citation_style"])
annotated_citations=[]
if newDocument.metadata["original_citation_style"]=="APA":
for index,citation in enumerate(citations_found):
newCit=newDocument.addCitation(sent_id=newSent["id"])
reference=matchCitationWithReference(citation, newDocument["references"])
## print (citation["text"]," -> ", formatReference(reference))
if reference:
newCit["ref_id"]=reference["id"]
else:
# do something else?
newCit["ref_id"]=None
annotated_citations.append(newCit)
annotated_s=replaceTempCitToken(annotated_s, index+1, newCit["id"])
elif newDocument.metadata["original_citation_style"]=="AFI":
for index,citation in enumerate(citations_found):
valid_citation=True
nums=[]
for num in citation["nums"]:
cit_num=int(num)-1
if cit_num < 0:
# this is not a citation! Probably something like "then sampled a random number from uniform distribution, u ~ U[0,1]"
valid_citation=False
break
nums.append(cit_num)
if not valid_citation:
continue
cit_ids=[]
for num in nums:
newCit=newDocument.addCitation(sent_id=newSent["id"])
# TODO check this: maybe not this simple? May need matching function.
newCit["ref_id"]="ref"+str(num)
cit_ids.append(newCit["id"])
annotated_citations.append(newCit)
annotated_s=replaceTempCitTokenMulti(annotated_s, index+1, cit_ids)
if len(annotated_citations) > 0:
newSent["citations"]=[acit["id"] for acit in annotated_citations]
newSent["text"]=annotated_s
