def cleanGhr():
fields = [
["title", [["id","tab1"]]], \
["reviewdate", [["class","datestamp","0"]]], \
["publishdate", [["class","datestamp","1"]]], \
["description", [["class","allcontent","0"]]]]
required = ([0,3], []) # lists for "mandatory" and "oneof"
doc_no = html2trec("../../data/ghr-data/", "ghr", "filelist", fields, required)
def cleanMadisons():
fields = [
["title", [["class","headerpaneopen","0"]]], \
["createdate", [["class","createdate","0"]]], \
["modifydate", [["class","modifydate","0"]]], \
["description", [["class","mpowerwrapper","0"]]]]
required = ([0,3], []) # lists for "mandatory" and "oneof"
doc_no = html2trec("../../data/madisons-data/", "madisons", "filelist", fields, required)
def cleanGard():
default_desc = 'These Web pages are updated as the Genetic and Rare Diseases Information Center receives questions and as new information becomes available.'
fields = [
["title", [["id","lblTitle"]]], \
["synonyms", [["id","divSynonymColumn"]]], \
["description", [["id","lblDescriptionQuestion"]], default_desc], \
["references", [["id","divReferences"]]]]
required = ([0], [1,2]) # lists for "mandatory" and "oneof"
doc_no = html2trec("../../data/gard-data/", "gard", "filelist", fields, required)
def cleanHon():
fields = [
["title", [["class","input","5"],["xpath","option","0"]]], \
["description", [["class","txtbeige","0"]]]]
required = ([0,1], []) # lists for "mandatory" and "oneof"
doc_no = html2trec("../../data/hon-data/", "hon", "filelist", fields, required)