def __init__(self):
#patterns
self._pat_abstract_1 = '(Abstract|ABSTRACT)'
self._pat_abstract_2 = '(In this|This (paper|study|article|report)'
self._pat_abstract_2 += '|IN THIS|THIS (PAPER|STUDY|ARTICLE|REPORT))'
self._pat_abstract_end = '(!A_E!|$|\n\s*\n)'
self._pat_keywords = 'I[nN][dD][eE][xX] ?[tT][eE][rR][mM][sS]|'
self._pat_keywords += 'K[eE][yY] ?[wW][oO][rR][dD][sS](?: and [A-Za-z]*?)?|'
self._pat_keywords += 'G[eE][nN][eE][rR][aA][lL] ?[tT][eE][rR][mM][sS]?'
self.pat_months = 'janu?a?r?y?\.?|febr?u?a?r?y?\.?|marc?h?\.?|apri?l?\.?|'
self.pat_months += 'may|june?\.?|july?\.?|augu?s?t?\.?|sept?e?m?b?e?r?\.?|'
self.pat_months += 'octo?b?e?r?\.?|nove?m?b?e?r?\.?|dece?m?b?e?r?\.?'
self.pat_junctions = 'the|be|a|an|anthe|of|in|at|for|from|to|into|and|or|with'
self.pat_middle_name = '(ter|Ter|van|den|der|de|di|la|van der|von|chen|'
self.pat_middle_name += 'van de|van den|Van|Den|Der|De|Di|La|Van der|Von|'
self.pat_middle_name += 'Chen|Van de|Van den|el|El)'
self.alone_name = \
'([eE]t[. ]*?al\.?|[Ss]ons?|[Jj]r[ .,]|[Jj]unior|[eE]tc\.?)'
self.roman_num = '(I[ .]|II[ .]|III[ .]|IV[ .]|V[ .]|VI[ .]|VII[ .]|'
self.roman_num += 'VIII[ .]|IX[ .]|X[ .])'
#regular expressions
self._re_abstract_1 = re.compile(self._pat_abstract_1 + '\W+(.+?)'
+ self._pat_abstract_end, re.DOTALL)
self._re_abstract_2 = re.compile(self._pat_abstract_2 + '\W+(.+?)'
+ self._pat_abstract_end, re.DOTALL)
self._re_keywords_1 = re.compile('(' + self._pat_keywords
+ ')[-:,;. ]*?\n+(.+?)\n')
self._re_keywords_2 = re.compile('(' + self._pat_keywords
+ ')[-:,;. ]+(.+?)(\n|[^0-9A-Z]\.|$)')
self._re_meta = re.compile('(^.*)(authors?:|emails?:|editors?:)', re.I)
self.re_published = re.compile('(^|\s)published (in|with)', re.I)
self.re_issn = re.compile('ISSN|[0-9]{4}-[0-9]{4}')
self.re_rep = re.compile('(^\s*[a-z]{2,} report\s*$|tech report)', re.I)
self.re_lower_start = re.compile('^[a-z][^A-Z-\'`]')
self.re_vol = re.compile('vol(ume)?\.? *[0-9]+', re.I)
self.re_rev = re.compile('[Rr]evision *[0-9]+')
self.re_inproc = \
re.compile('in proc\.|in proceedings|proceedings of', re.I)
self.re_no = re.compile('n(o|umber)\.? *[0-9]+', re.I)
self.re_pages = re.compile('(pages?|p\.) [0-9]', re.I)
self.re_etal = re.compile('et al\.?(\s|$)', re.I)
self.re_copyright = re.compile('(^|\s)copyright(\s|$)', re.I)
self.re_date = re.compile('(' + self.pat_months + ') .*?[0-9]{4}', re.I)
self.re_year = re.compile('\([0-9]{4}\)')
self.re_num = re.compile('^[0-9\s%]+$', re.DOTALL)
self.re_noletter = re.compile('^[^a-z]+$', re.IGNORECASE)
self.re_upper_word = re.compile('[A-Z][A-Z]+')
self.re_lower_word = re.compile('[a-z]')
self.re_telfax = re.compile('(tel|fax) +[+]?[0-9][-0-9 ()]', re.I)
self.re_empty = re.compile('^\s*$', re.DOTALL)
self.re_notitle = re.compile('\s(conference|symposium)\s', re.M | re.I)
self.re_title = re.compile('[A-Z][A-Z]+( [A-Z][A-Z])* [0-9]{4}')
self.re_mark = \
re.compile('(^[^\s]+ ?(/|-[0-9]) ?[^\s]+$|[0-9]+[-:][0-9]+)')
self.re_type = \
re.compile('(^|\s)thesis|article|journal(\s|$)', re.I | re.DOTALL)
self.re_organization = \
re.compile('department of|university|school of', re.IGNORECASE)
self.re_press = \
re.compile('(^|[^0-9a-z])in press([^0-9a-z]|$)', re.IGNORECASE)
self.re_domain = re.compile('.+\.[a-z]{2,3}$', re.IGNORECASE)
self.re_zav = re.compile('^[[(].*[])]$')
self.re_title = re.compile('^.*?T[Ii][Tt][Ll][Ee]:\s*(.+?)(\.|$)')
self.re_authors = re.compile('^[ ,.&]*('
'(([A-Z][-A-Za-z\'´`]+[.,]?|' + self.pat_middle_name +
'[.,]?|([A-Z]\.?-)?[A-Z]\.)[ ]+)*?'
'(' #Jmeno von Prijmeni
'(([A-Z][-A-Za-z\'´`]+)[.,]?[ ]+)'
'(([A-Z][-A-Za-z\'´`]+[.,]?|' + self.pat_middle_name +
'[.,]?|([A-Z]\.?-)?[A-Z]\.)[ ]+)*?'
'(([A-Z][-A-Za-z\'´`]+)[ ]*)'
'|' #Prijmeni, J.
'(([A-Z][-A-Za-z\'´`]+[.,]?)[ ]+)'
'(([A-Z][-A-Za-z\'´`]+[.,]?|' + self.pat_middle_name +
'[.,]?|([A-Z]\.?-)?[A-Z]\.)[ ]+)*?'
'((' + self.pat_middle_name
+ '[.,]?|([A-Z]\.?-)?[A-Z][ .]?)[ ]*)+?'
'|' #J. von Prijmeni
'((' + self.pat_middle_name
+ '[.,]?|([A-Z]\.?-)?[A-Z]\.)[ ]+)+?'
'(([A-Z][-A-Za-z\'´`]+[.,]?|' + self.pat_middle_name +
'[.,]?|([A-Z]\.?-)?[A-Z]\.)[ ]+)*?'
'(([A-Z][-A-Za-z\'´`]+)[ ]*)'
')'
'(([A-Z]v+|' + self.pat_middle_name
+ '|([A-Z]\.?-)?[A-Z]\.)[ ]*)*?'
'(([., ]*(' + self.alone_name + '|'
+ self.roman_num + ')?)*'
'((?<=\.) |&|[,. ]?and[,. ]|,|\.|;|$|%SEP%|\())+?'
'([., ]*' + self.alone_name + ')*)', re.VERBOSE)
self.re_inc = re.compile('[^a-zA-Z]([A-Z]\.|' + self.pat_middle_name
+ '|' + self.alone_name + '|&)[^a-zA-Z]')
self.re_end = re.compile('([-:,]| (' + self.pat_junctions + '))\s*\n',
re.IGNORECASE)
self.re_split = \
re.compile('^[A-Z]+( [^ a-z]+)*\n([^ a-z]+( [^ a-z]+)*\n)+')
self.re_upper = re.compile('([A-Z][A-Z]+ ){3}')
self.re_lower = re.compile('(^.+?) (.[^ ]?[a-z].*$)')
self.re_by = re.compile('(^.+) by (.+?)$', re.DOTALL)
self.re_autinline = \
re.compile('^([A-Z][^A-Z]+( [^A-Z]+){3}.*?)([A-Z][.a-z].*)')
self.re_word_end = re.compile(' [^A-Z]+\s*$')
self.re_term = re.compile('(^|\s*[:,;]+)\s*(.+?)\s*([:,;]+\s*|$)')
#Dictionaries
try:
self.rrsdict_locations = RRSDictionary(COUNTRIES, FIRST_UPPER)
self.rrsdict_locations.extend(RRSDictionary(CITIES, FIRST_UPPER))
except RRSDictionaryError:
raise DictionaryError("Failed to load dictionaries.")
def __init__(self):
self.pat_month = "january|february|march|april|may|june|july|august|"
self.pat_month += "september|october|november|december|jan\.?|feb\.?|"
self.pat_month += "mar\.?|apr\.?|jun\.?|jul\.?|aug\.?|sept?\.?|oct\.?|"
self.pat_month += "nov\.?|dec\.?"
self.pat_date = "(" + self.pat_month + \
" ([_0-9]+\.? )*)(19[0-9]{2}|2[0123][0-9]{2})"
self.kws_unpublished = {
"introduction": 50,
"abstract": 50,
"related work": 50
}
self.kws_article = {
"(vol\.?|volume)\s*[0-9]+": 20,
self.pat_date: 20,
"(pages?|pp?\.?)\s*[-0-9]+(?!\))": 20,
"(number|no\.?)\s*[_0-9]+": 20,
"copyright": 20,
"all rights reserved": 20,
"journal": 20,
"is published": 20,
"published in": 20,
"first published": 20,
"in press": 20,
"introduction": 50,
"abstract": 50,
"related work": 50,
self._find_events: 50,
self._find_events_2: 20
}
self.kws_techreport = {
"this report": 200,
"tech[a-z]+ report": 200,
"summary report": 100,
"is (a )?report": 80
}
self.kws_phdthesis = {
"supervisor": 100,
"this thesis": 200,
"dissertation": 200,
"Ph\.?D thesis": 200
}
self.kws_masterthesis = {
"supervisor": 100,
"this thesis": 200,
"master thesis": 200,
"master\W?s thesis": 200
}
self.types = {
UNPUBLISHED: self.kws_unpublished,
ARTICLE: self.kws_article,
PHDTHESIS: self.kws_phdthesis,
MASTERTHESIS: self.kws_masterthesis,
TECHREPORT: self.kws_techreport
}
self.re_proceedings = re.compile('\W(proceedings|conference)\W',
re.DOTALL)
self.re_ms_powerpoint = re.compile('powerpoint', re.IGNORECASE)
self.re_oo_impress = re.compile('impress', re.IGNORECASE)
self.re_pages = re.compile('Pages:\s*([0-9]+)')
self.pat_chapters = "R(eferences|EFERENCES)"
self.re_chapters = re.compile('(^.*\W)(' + self.pat_chapters + ')\W',
re.DOTALL)
#Proceedings articles patterns and RE
self.pat_time = "[0-2]?[0-9]:[0-9][0-9]"
self.pat_toc_page = "(\.\s*){2,}[0-9]+\n"
self.pat_pagesep = "(\s*\n\n\s*|\s*-\s*\n\n|\n\n\s*-\s*)"
self.pat_roman_nums = self.pat_pagesep + \
"(XX|XIX|XVIII|XVII|XVI|XV|XIV|XIII|XII|XI|X|IX|VIII|VII|VI|V|IV|III|II|I)" + \
self.pat_pagesep
self.pat_page_end_strict = "(" + self.pat_pagesep + "[0-9]+" + \
self.pat_pagesep + "|\npage [0-9]+\n|\.[0-9]+" + \
self.pat_pagesep + ")"
self.pat_page_end = "(" + self.pat_pagesep + "[0-9]+" + \
self.pat_pagesep + "|\npage [0-9]+\n|\.[0-9]+\n\n|\n\n)"
self.re_proceedings_prefix_strict = \
re.compile("(^.+((" + self.pat_time + "|" + self.pat_toc_page + \
").*" + self.pat_page_end_strict + "|" + \
self.pat_roman_nums + "))", re.DOTALL | re.IGNORECASE)
self.re_proceedings_prefix = \
re.compile("(^.+((" + self.pat_time + "|" + self.pat_toc_page + \
").*" + self.pat_page_end + "|" + self.pat_roman_nums + \
"))", re.DOTALL | re.IGNORECASE)
self.re_abstract = \
re.compile("(^.+?)\n(abstract|references)(\W[^.\n]*)?\n",
re.DOTALL | re.IGNORECASE)
self.re_proceedings = re.compile("(^.+?\Wproceedings\W.*?\n\n)",
re.DOTALL | re.IGNORECASE)
self.re_multi_strict = \
re.compile('(^.{5000,}?\nR[Ee][Ff][Ee][Rr][Ee][Nn][Cc][Ee][Ss]\W*\n.+?' + \
self.pat_page_end_strict + ')', re.DOTALL)
self.re_previous_1 = \
re.compile('(^\s*([A-Z]\.|[A-Z]\w+[,. ]+[A-Z]\.|' + \
'[^\n]+?[(.,]\s*[0-9]{4}\s*[.),][^\n]+?\n|' + \
'[^a-zA-Z]|[a-z]|[A-Z]\w*?\.?\n|[^\n]*\Weds\W).*?' + \
self.pat_page_end + ')', re.DOTALL)
self.re_previous_2 = \
re.compile('^\s*(Introduction|INTRODUCTION|Appendix|APPENDIX|' + \
'Table\W+[0-9]|TABLE\W+[0-9]|Figure\W+[0-9]|' + \
'FIGURE\W+[0-9]|Section\W+[0-9]|SECTION\W+[0-9])',
re.DOTALL)
self.rrsdict_events = RRSDictionary(EVENT_ACRONYMS, CASE_SENSITIVE)
