Grammalecte  Check-in [3ef2bdb736]

Overview
Comment:[graphspell] tokenizer: combining diacritics recognition and NFC normalization
Downloads: Tarball | ZIP archive | SQL archive
Timelines: family | ancestors | descendants | both | trunk | graphspell
Files: files | file ages | folders
SHA3-256: 3ef2bdb736e458181f90ed3aef50d68cba41471e0f3d6090a05cf4ef03128595
User & Date: olr on 2020-04-20 18:02:29
Other Links: manifest | tags
Context
2020-04-21
08:11
[fr] ajustements check-in: 014e846ccc user: olr tags: fr, trunk, v1.9.0
2020-04-20
18:02
[graphspell] tokenizer: combining diacritics recognition and NFC normalization check-in: 3ef2bdb736 user: olr tags: graphspell, trunk
17:44
[fr] ajustements check-in: 04d726b729 user: olr tags: fr, trunk
Changes

Modified gc_lang/fr/perf_memo.txt from [70dc62f0cc] to [9508c4062f].

24
25
26
27
28
29
30


0.5.16      2017.05.12 07:41    4.92201     1.19269     0.80639     0.239147    0.257518    0.266523    0.62111     0.33359     0.0634668   0.00757178
0.6.1       2018.02.12 09:58    5.25924     1.2649      0.878442    0.257465    0.280558    0.293903    0.686887    0.391275    0.0672474   0.00824723
0.6.2       2018.02.19 19:06    5.51302     1.29359     0.874157    0.260415    0.271596    0.290641    0.684754    0.376905    0.0815201   0.00919633  (spelling normalization)
1.0         2018.11.23 10:59    2.88577     0.702486    0.485648    0.139897    0.14079     0.148125    0.348751    0.201061    0.0360297   0.0043535   (x2, with new GC engine)
1.1         2019.05.16 09:42    1.50743     0.360923    0.261113    0.0749272   0.0763827   0.0771537   0.180504    0.102942    0.0182762   0.0021925   (×2, but new processor: AMD Ryzen 7 2700X)
1.2.1       2019.08.06 20:57    1.42886     0.358425    0.247356    0.0704405   0.0754886   0.0765604   0.177197    0.0988517   0.0188103   0.0020243
1.6.0       2020.01.03 20:22    1.38847     0.346214    0.240242    0.0709539   0.0737499   0.0748733   0.176477    0.0969171   0.0187857   0.0025143   (nouveau dictionnaire avec lemmes masculin)









>
>
24
25
26
27
28
29
30
31
32
0.5.16      2017.05.12 07:41    4.92201     1.19269     0.80639     0.239147    0.257518    0.266523    0.62111     0.33359     0.0634668   0.00757178
0.6.1       2018.02.12 09:58    5.25924     1.2649      0.878442    0.257465    0.280558    0.293903    0.686887    0.391275    0.0672474   0.00824723
0.6.2       2018.02.19 19:06    5.51302     1.29359     0.874157    0.260415    0.271596    0.290641    0.684754    0.376905    0.0815201   0.00919633  (spelling normalization)
1.0         2018.11.23 10:59    2.88577     0.702486    0.485648    0.139897    0.14079     0.148125    0.348751    0.201061    0.0360297   0.0043535   (x2, with new GC engine)
1.1         2019.05.16 09:42    1.50743     0.360923    0.261113    0.0749272   0.0763827   0.0771537   0.180504    0.102942    0.0182762   0.0021925   (×2, but new processor: AMD Ryzen 7 2700X)
1.2.1       2019.08.06 20:57    1.42886     0.358425    0.247356    0.0704405   0.0754886   0.0765604   0.177197    0.0988517   0.0188103   0.0020243
1.6.0       2020.01.03 20:22    1.38847     0.346214    0.240242    0.0709539   0.0737499   0.0748733   0.176477    0.0969171   0.0187857   0.0025143   (nouveau dictionnaire avec lemmes masculin)
1.9.0       2020.04.20 19:57    1.51183     0.369546    0.25681     0.0734314   0.0764396   0.0785668   0.183922    0.103674    0.0185812   0.002099    (NFC normalization)

Modified graphspell-js/tokenizer.js from [16e7826100] to [efabea9cdf].

40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
..
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
            [/^\[\/?[a-zA-Z]+\]/, 'PSEUDOHTML'],
            [/^&\w+;(?:\w+;|)/, 'HTMLENTITY'],
            [/^(?:l|d|n|m|t|s|j|c|ç|lorsqu|puisqu|jusqu|quoiqu|qu|presqu|quelqu)['’´‘′`ʼ]/i, 'WORD_ELIDED'],
            [/^\d\d?[h:]\d\d(?:[m:]\d\ds?|)\b/, 'HOUR'],
            [/^\d+(?:ers?\b|res?\b|è[rm]es?\b|i[èe][mr]es?\b|de?s?\b|nde?s?\b|ès?\b|es?\b|ᵉʳˢ?|ʳᵉˢ?|ᵈᵉ?ˢ?|ⁿᵈᵉ?ˢ?|ᵉˢ?)/, 'WORD_ORDINAL'],
            [/^\d+(?:[.,]\d+|)/, 'NUM'],
            [/^[&%‰€$+±=*/<>⩾⩽#|×¥£§¢¬÷@-]/, 'SIGN'],
            [/^[a-zA-Zà-öÀ-Ö0-9ø-ÿØ-ßĀ-ʯff-stᴀ-ᶿᵉʳˢⁿᵈ_]+(?:[’'`-][a-zA-Zà-öÀ-Ö0-9ø-ÿØ-ßĀ-ʯff-stᴀ-ᶿᵉʳˢⁿᵈ_]+)*/, 'WORD']
        ]
};


class Tokenizer {

    constructor (sLang) {
................................................................................
        while (sText) {
            let iCut = 1;
            for (let [zRegex, sType] of this.aRules) {
                if (sType !== "SPACE"  ||  bWithSpaces) {
                    try {
                        if ((m = zRegex.exec(sText)) !== null) {
                            iToken += 1;
                            yield { "i": iToken, "sType": sType, "sValue": m[0], "nStart": iNext, "nEnd": iNext + m[0].length };
                            iCut = m[0].length;
                            break;
                        }
                    }
                    catch (e) {
                        console.error(e);
                    }







|







 







|







40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
..
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
            [/^\[\/?[a-zA-Z]+\]/, 'PSEUDOHTML'],
            [/^&\w+;(?:\w+;|)/, 'HTMLENTITY'],
            [/^(?:l|d|n|m|t|s|j|c|ç|lorsqu|puisqu|jusqu|quoiqu|qu|presqu|quelqu)['’´‘′`ʼ]/i, 'WORD_ELIDED'],
            [/^\d\d?[h:]\d\d(?:[m:]\d\ds?|)\b/, 'HOUR'],
            [/^\d+(?:ers?\b|res?\b|è[rm]es?\b|i[èe][mr]es?\b|de?s?\b|nde?s?\b|ès?\b|es?\b|ᵉʳˢ?|ʳᵉˢ?|ᵈᵉ?ˢ?|ⁿᵈᵉ?ˢ?|ᵉˢ?)/, 'WORD_ORDINAL'],
            [/^\d+(?:[.,]\d+|)/, 'NUM'],
            [/^[&%‰€$+±=*/<>⩾⩽#|×¥£§¢¬÷@-]/, 'SIGN'],
            [/^[a-zA-Zà-öÀ-Ö0-9ø-ÿØ-ßĀ-ʯff-stᴀ-ᶿ\u0300-\u036fᵉʳˢⁿᵈ_]+(?:[’'`-][a-zA-Zà-öÀ-Ö0-9ø-ÿØ-ßĀ-ʯff-stᴀ-ᶿ\u0300-\u036fᵉʳˢⁿᵈ_]+)*/, 'WORD']
        ]
};


class Tokenizer {

    constructor (sLang) {
................................................................................
        while (sText) {
            let iCut = 1;
            for (let [zRegex, sType] of this.aRules) {
                if (sType !== "SPACE"  ||  bWithSpaces) {
                    try {
                        if ((m = zRegex.exec(sText)) !== null) {
                            iToken += 1;
                            yield { "i": iToken, "sType": sType, "sValue": m[0], "nStart": iNext, "nEnd": iNext + m[0].length };  // m[0].normalize("NFC") not usefull at the moment
                            iCut = m[0].length;
                            break;
                        }
                    }
                    catch (e) {
                        console.error(e);
                    }

Modified graphspell/tokenizer.py from [a2c42f5f3e] to [81da836011].

1
2
3
4
5
6
7


8
9
10
11
12
13
14
..
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
..
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
"""
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<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|presqu|quelqu)['’´‘′`ʼ])",
            r'(?P<WORD_ORDINAL>\d+(?:ers?|res?|è[rm]es?|i[èe][mr]es?|de?s?|nde?s?|ès?|es?|ᵉʳˢ?|ʳᵉˢ?|ᵈᵉ?ˢ?|ⁿᵈᵉ?ˢ?|ᵉˢ?)\b)',
            r'(?P<HOUR>\d\d?[h:]\d\d(?:[m:]\d\ds?|)\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 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>"] }

    def getTokenTypes (self):
        "returns list of token types as tuple (token name, regex)"
        return [ sRegex[4:-1].split(">")  for sRegex in _PATTERNS[self.sLang] ]







>
>







 







|







 







|







1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
..
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
..
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
"""
Very simple tokenizer
using regular expressions
"""

import re

from unicodedata import normalize

_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<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|presqu|quelqu)['’´‘′`ʼ])",
            r'(?P<WORD_ORDINAL>\d+(?:ers?|res?|è[rm]es?|i[èe][mr]es?|de?s?|nde?s?|ès?|es?|ᵉʳˢ?|ʳᵉˢ?|ᵈᵉ?ˢ?|ⁿᵈᵉ?ˢ?|ᵉˢ?)\b)',
            r'(?P<HOUR>\d\d?[h:]\d\d(?:[m:]\d\ds?|)\b)',
            r'(?P<NUM>\d+(?:[.,]\d+|))',
            r'(?P<SIGN>[&%‰€$+±=*/<>⩾⩽#|×¥£¢§¬÷@-])',
            r"(?P<WORD>[\w\u0300-\u036f]+(?:[’'`-][\w\u0300-\u036f]+)*)"
        )
}


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

................................................................................

    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": normalize("NFC", 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>"] }

    def getTokenTypes (self):
        "returns list of token types as tuple (token name, regex)"
        return [ sRegex[4:-1].split(">")  for sRegex in _PATTERNS[self.sLang] ]