Grammalecte  Check-in [1edae62ad8]

Overview
Comment:[core] ibdawg: another suggestion method
Downloads: Tarball | ZIP archive | SQL archive
Timelines: family | ancestors | descendants | both | core | spellsugg
Files: files | file ages | folders
SHA3-256: 1edae62ad8b792d29eba111b042aee28860ba0174487324ef0bf92ebc40dc7fe
User & Date: olr on 2017-11-07 11:56:37
Other Links: branch diff | manifest | tags
Context
2017-11-07
17:03
[core] ibdawg: use SuggResult object + code cleaning check-in: 6c8c1776f5 user: olr tags: core, spellsugg
11:56
[core] ibdawg: another suggestion method check-in: 1edae62ad8 user: olr tags: core, spellsugg
2017-11-06
17:21
[core][py] timer for testing check-in: 8ddc15536e user: olr tags: trunk, core
Changes

Modified cli.py from [ed7458a4d3] to [a5ccfdebb3].

206
207
208
209
210
211
212

213
214
215
216
217
218
219
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220







+







                        echo("* " + sWord)
                        for sMorph in oDict.getMorph(sWord):
                            echo("  {:<32} {}".format(sMorph, oLexGraphe.formatTags(sMorph)))
            elif sText.startswith("!"):
                for sWord in sText[1:].strip().split():
                    if sWord:
                        echo(" | ".join(oDict.suggest(sWord)))
                        echo(" | ".join(oDict.suggest2(sWord)))
            elif sText.startswith(">"):
                oDict.drawPath(sText[1:].strip())
            elif sText.startswith("="):
                for sRes in oDict.select(sText[1:].strip()):
                    echo(sRes)
            elif sText.startswith("/+ "):
                gce.setOptions({ opt:True  for opt in sText[3:].strip().split()  if opt in gce.getOptions() })

Modified gc_core/py/char_player.py from [97f69bd70d] to [19df5311c8].

44
45
46
47
48
49
50





























































51
52
53
54
55
56
57
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118







+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+








def showDistance (s1, s2):
    s1b = cleanWord(s1);
    s2b = cleanWord(s2);
    print(f"Distance: {s1} / {s2} = {distanceDamerauLevenshtein(s1, s2)}")
    print(f"Distance: {s1b} / {s2b} = {distanceDamerauLevenshtein(s1b, s2b)}")


def distanceSift4 (s1, s2, nMaxOffset=5):
    "implementation of general Sift4."
    # https://siderite.blogspot.com/2014/11/super-fast-and-accurate-string-distance.html
    if not s1:
        return len(s2)
    if not s2:
        return len(s1)
    nLen1, nLen2 = len(s1), len(s2)
    i1, i2 = 0, 0   # Cursors for each string
    nLargestCS = 0  # Largest common substring
    nLocalCS = 0    # Local common substring
    nTrans = 0      # Number of transpositions ('ab' vs 'ba')
    lOffset = []    # Offset pair array, for computing the transpositions
 
    while i1 < nLen1 and i2 < nLen2:
        if s1[i1] == s2[i2]:
            nLocalCS += 1
            # Check if current match is a transposition
            bTrans = False
            i = 0
            while i < len(lOffset):
                t = lOffset[i]
                if i1 <= t[0] or i2 <= t[1]:
                    bTrans = abs(i2-i1) >= abs(t[1] - t[0])
                    if bTrans:
                        nTrans += 1
                    elif not t[2]:
                        t[2] = True
                        nTrans += 1
                    break
                elif i1 > t[1] and i2 > t[0]:
                    del lOffset[i]
                else:
                    i += 1
            lOffset.append([i1, i2, bTrans])
        else:
            nLargestCS += nLocalCS
            nLocalCS = 0
            if i1 != i2:
                i1 = i2 = min(i1, i2)
            for i in range(nMaxOffset):
                if i1 + i >= nLen1 and i2 + i >= nLen2:
                    break
                elif i1 + i < nLen1 and s1[i1+i] == s2[i2]:
                    i1 += i - 1
                    i2 -= 1
                    break
                elif i2 + i < nLen2 and s1[i1] == s2[i2+i]:
                    i2 += i - 1
                    i1 -= 1
                    break
        i1 += 1
        i2 += 1
        if i1 >= nLen1 or i2 >= nLen2:
            nLargestCS += nLocalCS
            nLocalCS = 0
            i1 = i2 = min(i1, i2)
    nLargestCS += nLocalCS
    return round(max(nLen1, nLen2) - nLargestCS + nTrans)


# Method: Remove Useless Chars

_dVovels = {
    'a': '',  'e': '',  'i': '',  'o': '',  'u': '',  'y': '',
    'à': '',  'é': '',  'î': '',  'ô': '',  'û': '',  'ÿ': '',
    'â': '',  'è': '',  'ï': '',  'ö': '',  'ù': '',  'ŷ': '',

Modified gc_core/py/ibdawg.py from [a0b40e49a1] to [b7c1b974f3].

1
2
3
4
5
6
7
8
9
10
11
12


13
14
15
16
17
18
19
1
2
3
4
5
6
7
8
9
10


11
12
13
14
15
16
17
18
19










-
-
+
+







#!python3

import os
import traceback
import pkgutil
import re
from itertools import chain
from functools import wraps
import time

#import logging
#logging.basicConfig(filename="suggestions.log", level=logging.DEBUG)
import logging
logging.basicConfig(filename="suggestions.log", level=logging.DEBUG)

from . import str_transform as st
from . import char_player as cp
from .echo import echo


def timethis (func):
216
217
218
219
220
221
222
223

224
225
226

227
228
229
230
231
232
233
234
235
236
237
238

239
240
241
242

243
244
245

246
247
248

249
250
251

252
253

254
255
256
257
258
259
260

261
262
263
264

265
266

267
268

269
270









































271
272
273
274
275
276
277
216
217
218
219
220
221
222

223
224
225

226
227
228
229
230
231
232
233
234
235
236
237

238
239
240
241

242
243
244

245
246
247

248
249
250

251
252

253
254
255
256
257
258
259

260
261
262
263

264
265

266
267

268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318







-
+


-
+











-
+



-
+


-
+


-
+


-
+

-
+






-
+



-
+

-
+

-
+


+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+







        sCleanWord = cp.cleanWord(sWord)
        aSugg = sorted(aSugg, key=lambda sSugg: cp.distanceDamerauLevenshtein(sCleanWord, cp.cleanWord(sSugg)))[:nMaxSugg]
        if sSfx or sPfx:
            # we add what we removed
            return list(map(lambda sSug: sPfx + sSug + sSfx, aSugg))
        return aSugg

    def _suggest (self, sRemain, nMaxDel=0, nMaxHardRepl=0, nDeep=0, iAddr=0, sNewWord="", bAvoidLoop=False):
    def _suggest (self, sRemain, nMaxDel=0, nMaxHardRepl=0, nDeep=0, iAddr=0, sNewWord="", sAction="", bAvoidLoop=False):
        "returns a set of suggestions"
        # recursive function
        #logging.info((nDeep * "  ") + sNewWord + ":" + sRemain)
        #logging.info((nDeep * "  ") + sNewWord + ":" + sRemain + " · " + sAction)
        aSugg = set()
        if not sRemain:
            if int.from_bytes(self.byDic[iAddr:iAddr+self.nBytesArc], byteorder='big') & self._finalNodeMask:
                #logging.info((nDeep * "  ") + "__" + sNewWord + "__")
                aSugg.add(sNewWord)
            for sTail in self._getTails(iAddr):
                #logging.info((nDeep * "  ") + "__" + sNewWord+sTail + "__")
                aSugg.add(sNewWord+sTail)
            return aSugg
        cCurrent = sRemain[0:1]
        for cChar, jAddr in self._getSimilarArcs(cCurrent, iAddr):
            aSugg.update(self._suggest(sRemain[1:], nMaxDel, nMaxHardRepl, nDeep+1, jAddr, sNewWord+cChar))
            aSugg.update(self._suggest(sRemain[1:], nMaxDel, nMaxHardRepl, nDeep+1, jAddr, sNewWord+cChar, "*"))
        if not bAvoidLoop: # avoid infinite loop
            if cCurrent == sRemain[1:2]:
                # same char, we remove 1 char without adding 1 to <sNewWord>
                aSugg.update(self._suggest(sRemain[1:], nMaxDel, nMaxHardRepl, nDeep+1, iAddr, sNewWord))
                aSugg.update(self._suggest(sRemain[1:], nMaxDel, nMaxHardRepl, nDeep+1, iAddr, sNewWord, cCurrent+"/2"))
            else:
                # switching chars
                aSugg.update(self._suggest(sRemain[1:2]+sRemain[0:1]+sRemain[2:], nMaxDel, nMaxHardRepl, nDeep+1, iAddr, sNewWord, True))
                aSugg.update(self._suggest(sRemain[1:2]+sRemain[0:1]+sRemain[2:], nMaxDel, nMaxHardRepl, nDeep+1, iAddr, sNewWord, "><",True))
                # delete char
                if nMaxDel > 0:
                    aSugg.update(self._suggest(sRemain[1:], nMaxDel-1, nMaxHardRepl, nDeep+1, iAddr, sNewWord, True))
                    aSugg.update(self._suggest(sRemain[1:], nMaxDel-1, nMaxHardRepl, nDeep+1, iAddr, sNewWord, "-"+cCurrent, True))
            # Phonetic replacements
            for sRepl in cp.d1toX.get(cCurrent, ()):
                aSugg.update(self._suggest(sRepl + sRemain[1:], nMaxDel, nMaxHardRepl, nDeep+1, iAddr, sNewWord, True))
                aSugg.update(self._suggest(sRepl + sRemain[1:], nMaxDel, nMaxHardRepl, nDeep+1, iAddr, sNewWord, cCurrent+">"+sRepl, True))
            for sRepl in cp.d2toX.get(sRemain[0:2], ()):
                aSugg.update(self._suggest(sRepl + sRemain[2:], nMaxDel, nMaxHardRepl, nDeep+1, iAddr, sNewWord, True))
                aSugg.update(self._suggest(sRepl + sRemain[2:], nMaxDel, nMaxHardRepl, nDeep+1, iAddr, sNewWord, sRemain[0:2]+">"+sRepl, True))
            # Hard replacements
            if nDeep > 3 and nMaxHardRepl and len(sRemain) >= 2:
                for nVal, kAddr in self._getArcs1(iAddr):
                    if nVal in self.dCharVal:
                        cChar = self.dCharVal[nVal]
                        if cChar not in cp.d1to1.get(cCurrent, ""):
                            aSugg.update(self._suggest(sRemain[1:], nMaxDel, nMaxHardRepl-1, nDeep+1, kAddr, sNewWord+cChar, True))
                            aSugg.update(self._suggest(sRemain[1:], nMaxDel, nMaxHardRepl-1, nDeep+1, kAddr, sNewWord+cChar, "[["+cChar+"]]", True))
            # end of word
            if len(sRemain) == 2:
                for sRepl in cp.dFinal2.get(sRemain, ()):
                    aSugg.update(self._suggest(sRepl, nMaxDel, nMaxHardRepl, nDeep+1, iAddr, sNewWord, True))
                    aSugg.update(self._suggest(sRepl, nMaxDel, nMaxHardRepl, nDeep+1, iAddr, sNewWord, sRemain + " >> " + sRepl, True))
            elif len(sRemain) == 1:
                aSugg.update(self._suggest("", nMaxDel, nMaxHardRepl, nDeep+1, iAddr, sNewWord, True)) # remove last char and go on
                aSugg.update(self._suggest("", nMaxDel, nMaxHardRepl, nDeep+1, iAddr, sNewWord, sRemain + " [last char removed] ", True)) # remove last char and go on
                for sRepl in cp.dFinal1.get(sRemain, ()):
                    aSugg.update(self._suggest(sRepl, nMaxDel, nMaxHardRepl, nDeep+1, iAddr, sNewWord, True))
                    aSugg.update(self._suggest(sRepl, nMaxDel, nMaxHardRepl, nDeep+1, iAddr, sNewWord, sRemain + " >> " + sRepl, True))
        return aSugg

    @timethis
    def suggest2 (self, sWord, nMaxSugg=10):
        "returns a set of suggestions for <sWord>"
        sPfx, sWord, sSfx = cp.cut(sWord)
        nMaxDist = (len(sWord) // 3) + 1
        sCleanWord = cp.cleanWord(sWord)
        aSugg = self._suggest2(sWord, sCleanWord, nMaxDist)
        if sWord.istitle():
            aSugg.update(self._suggest2(sWord.lower(), sCleanWord, nMaxDist))
            aSugg = set(map(lambda sSugg: sSugg.title(), aSugg))
        elif sWord.islower():
            aSugg.update(self._suggest2(sWord.title(), sCleanWord, nMaxDist))
        aSugg = cp.filterSugg(aSugg)
        aSugg = sorted(aSugg, key=lambda sSugg: cp.distanceSift4(sCleanWord, cp.cleanWord(sSugg)))[:nMaxSugg]
        if sSfx or sPfx:
            # we add what we removed
            return list(map(lambda sSug: sPfx + sSug + sSfx, aSugg))
        return aSugg

    def _suggest2 (self, sWord, sCleanWord, nMaxDist, nDeep=0, iAddr=0, sNewWord=""):
        #logging.info((nDeep * "  ") + sNewWord)
        if nDeep >= nMaxDist:
            sCleanNewWord = cp.cleanWord(sNewWord)
            if cp.distanceSift4(sCleanWord[:len(sCleanNewWord)], sCleanNewWord) > nMaxDist:
                return set()
        aSugg = set()
        if int.from_bytes(self.byDic[iAddr:iAddr+self.nBytesArc], byteorder='big') & self._finalNodeMask:
            #logging.info((nDeep * "  ") + "__" + sNewWord + "__")
            sCleanNewWord = cp.cleanWord(sNewWord)
            if cp.distanceSift4(sCleanWord, sCleanNewWord) <= nMaxDist:
                aSugg.add(sNewWord)
        for cChar, jAddr in self._getCharArcs(iAddr):
            aSugg.update(self._suggest2(sWord, sCleanWord, nMaxDist, nDeep+1, jAddr, sNewWord+cChar))
        return aSugg

    def _getCharArcs (self, iAddr):
        "generator: yield all chars and addresses from node at address <iAddr>"
        for nVal, jAddr in self._getArcs(iAddr):
            if nVal < self.nChar:
                yield (self.dCharVal[nVal], jAddr)

    def _getSimilarArcs (self, cChar, iAddr):
        "generator: yield similar char of <cChar> and address of the following node"
        for c in cp.d1to1.get(cChar, [cChar]):
            if c in self.dChar:
                jAddr = self._lookupArcNode(self.dChar[c], iAddr)
                if jAddr:
                    yield (c, jAddr)