Grammalecte  Artifact Content

Artifact daca54adb9dcf7cf8dac6f0425ce5ee792bc8b020d2f374496b304142229f8cb:

  • File graphspell/tokenizer.py — part of check-in [6950f5898f] at 2018-07-17 06:42:09 on branch rg — [graphspell] tokenizer: remove hyphen in number detection (always considered as a separate sign) (user: olr size: 2835) [more...]

"""
Very simple tokenizer
using regular expressions
"""

import re

_PATTERNS = {
    "default":
        (
            r'(?P<FOLDERUNIX>/(?:bin|boot|dev|etc|home|lib|mnt|opt|root|sbin|tmp|usr|var|Bureau|Documents|Images|Musique|Public|Téléchargements|Vidéos)(?:/[\w.()-]+)*)',
            r'(?P<FOLDERWIN>[a-zA-Z]:\\(?:Program Files(?: [(]x86[)]|)|[\w.()]+)(?:\\[\w.()-]+)*)',
            r'(?P<PUNC>[][,.;:!?…«»“”‘’"(){}·–—])',
            r'(?P<WORD_ACRONYM>[A-Z][.][A-Z][.](?:[A-Z][.])*)',
            r'(?P<LINK>(?:https?://|www[.]|\w+[@.]\w\w+[@.])\w[\w./?&!%=+*"\'@$#-]+)',
            r'(?P<HASHTAG>[#@][\w-]+)',
            r'(?P<HTML><\w+.*?>|</\w+ *>)',
            r'(?P<PSEUDOHTML>\[/?\w+\])',
            r'(?P<HOUR>\d\d?h\d\d\b)',
            r'(?P<NUM>\d+(?:[.,]\d+))',
            r'(?P<SIGN>[%‰+=*/<>⩾⩽-])',
            r"(?P<WORD>\w+(?:[’'`-]\w+)*)"
        ),
    "fr":
        (
            r'(?P<FOLDERUNIX>/(?:bin|boot|dev|etc|home|lib|mnt|opt|root|sbin|tmp|usr|var|Bureau|Documents|Images|Musique|Public|Téléchargements|Vidéos)(?:/[\w.()-]+)*)',
            r'(?P<FOLDERWIN>[a-zA-Z]:\\(?:Program Files(?: [(]x86[)]|)|[\w.()]+)(?:\\[\w.()-]+)*)',
            r'(?P<PUNC>[][,.;:!?…«»“”‘’"(){}·–—])',
            r'(?P<WORD_ACRONYM>[A-Z][.][A-Z][.](?:[A-Z][.])*)',
            r'(?P<LINK>(?:https?://|www[.]|\w+[@.]\w\w+[@.])\w[\w./?&!%=+*"\'@$#-]+)',
            r'(?P<HASHTAG>[#@][\w-]+)',
            r'(?P<HTML><\w+.*?>|</\w+ *>)',
            r'(?P<PSEUDOHTML>\[/?\w+\])',
            r"(?P<WORD_ELIDED>(?:l|d|n|m|t|s|j|c|ç|lorsqu|puisqu|jusqu|quoiqu|qu)['’`])",
            r'(?P<WORD_ORDINAL>\d+(?:ers?|nds?|es?|des?|ièmes?|èmes?|emes?|ᵉʳˢ?|ⁿᵈˢ?|ᵉˢ?|ᵈᵉˢ?)\b)',
            r'(?P<HOUR>\d\d?h\d\d\b)',
            r'(?P<NUM>\d+(?:[.,]\d+|))',
            r'(?P<SIGN>[%‰+=*/<>⩾⩽-])',
            r"(?P<WORD>\w+(?:[’'`-]\w+)*)"
        )
}


class Tokenizer:
    "Tokenizer: transforms a text in a list of tokens"

    def __init__ (self, sLang):
        self.sLang = sLang
        if sLang not in _PATTERNS:
            self.sLang = "default"
        self.zToken = re.compile( "(?i)" + '|'.join(sRegex for sRegex in _PATTERNS[sLang]) )

    def genTokens (self, sText, bStartEndToken=False):
        "generator: tokenize <sText>"
        i = 0
        if bStartEndToken:
            yield { "i": 0, "sType": "INFO", "sValue": "<start>", "nStart": 0, "nEnd": 0, "lMorph": ["<start>"] }
        for i, m in enumerate(self.zToken.finditer(sText), 1):
            yield { "i": i, "sType": m.lastgroup, "sValue": m.group(), "nStart": m.start(), "nEnd": m.end() }
        if bStartEndToken:
            iEnd = len(sText)
            yield { "i": i+1, "sType": "INFO", "sValue": "<end>", "nStart": iEnd, "nEnd": iEnd, "lMorph": ["<end>"] }