File size: 9,491 Bytes
ac901c7
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
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
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
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
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
#
#  Copyright (c) 2013-present, Anoop Kunchukuttan
#  All rights reserved.
#
#  This source code is licensed under the MIT license found in the
#  LICENSE file in the root directory of this source tree.
#

# Program for sentence splitting of Indian language input
#
# @author Anoop Kunchukuttan
#
"""

Sentence splitter for Indian languages. Contains a rule-based 

sentence splitter that can understand common non-breaking phrases

in many Indian languages.

"""

import regex as re
from indicnlp.transliterate import unicode_transliterate
from indicnlp import langinfo


## for language which have danda as delimiter
## period is not part of the sentence delimiters
DELIM_PAT_DANDA = re.compile(r"[\?!\u0964\u0965]")

## for languages which don't have danda as delimiter
DELIM_PAT_NO_DANDA = re.compile(
    r"[\.\?!\u0964\u0965\uAAF1\uAAF0\uABEB\uABEC\uABED\uABEE\uABEF\u1C7E\u1C7F]"
)

## pattern to check for presence of danda in text
CONTAINS_DANDA = re.compile(r"[\u0964\u0965]")

## pattern to check for presence of valid domain characters in text
CONTAINS_VALID_DOMAIN_CHAR = re.compile(r"^[a-zA-Z0-9_-]$")

## pattern to check for presence of multiple consecutive spaces in text
CONTAINS_MULTIPLE_SPACES = re.compile(" +")


def is_latin_or_numeric(character):
    """

    Check if a character is a Latin character (uppercase or lowercase) or a number.



    Parameters:

        character (str): The character to be checked.



    Returns:

        bool: True if the character is a Latin character or a number, False otherwise.

    """
    return re.match(CONTAINS_VALID_DOMAIN_CHAR, character) is not None


def is_acronym_abbvr(text, lang):
    """Is the text a non-breaking phrase



    Args:

        text (str): text to check for non-breaking phrase

        lang (str): ISO 639-2 language code



    Returns:

        boolean: true if `text` is a non-breaking phrase

    """

    ack_chars = {
        ## acronym for latin characters
        "ए",
        "ऎ",
        "बी",
        "बि",
        "सी",
        "सि",
        "डी",
        "डि",
        "ई",
        "इ",
        "एफ",
        "ऎफ",
        "जी",
        "जि",
        "एच",
        "ऎच",
        "आई",
        "आइ",
        "ऐ",
        "जे",
        "जॆ",
        "के",
        "कॆ",
        "एल",
        "ऎल",
        "एम",
        "ऎम",
        "एन",
        "ऎन",
        "ओ",
        "ऒ",
        "पी",
        "पि",
        "क्यू",
        "क्यु",
        "आर",
        "एस",
        "ऎस",
        "टी",
        "टि",
        "यू",
        "यु",
        "वी",
        "वि",
        "व्ही",
        "व्हि",
        "डब्ल्यू",
        "डब्ल्यु",
        "एक्स",
        "ऎक्स",
        "वाय",
        "जेड",
        "ज़ेड",
        ##  add halant to the previous English character mappings.
        "एफ्",
        "ऎफ्",
        "एच्",
        "ऎच्",
        "एल्",
        "ऎल्",
        "एम्",
        "ऎम्",
        "एन्",
        "ऎन्",
        "आर्",
        "एस्",
        "ऎस्",
        "एक्स्",
        "ऎक्स्",
        "वाय्",
        "जेड्",
        "ज़ेड्",
        # Indic vowels
        "ऄ",
        "अ",
        "आ",
        "इ",
        "ई",
        "उ",
        "ऊ",
        "ऋ",
        "ऌ",
        "ऍ",
        "ऎ",
        "ए",
        "ऐ",
        "ऑ",
        "ऒ",
        "ओ",
        "औ",
        "ॠ",
        "ॡ",
        # Indic consonants
        "क",
        "ख",
        "ग",
        "घ",
        "ङ",
        "च",
        "छ",
        "ज",
        "झ",
        "ञ",
        "ट",
        "ठ",
        "ड",
        "ढ",
        "ण",
        "त",
        "थ",
        "द",
        "ध",
        "न",
        "ऩ",
        "प",
        "फ",
        "ब",
        "भ",
        "म",
        "य",
        "र",
        "ऱ",
        "ल",
        "ळ",
        "ऴ",
        "व",
        "श",
        "ष",
        "स",
        "ह",
        ## abbreviation
        "श्री",
        "डॉ",
        "कु",
        "चि",
        "सौ",
    }

    return (
        unicode_transliterate.UnicodeIndicTransliterator.transliterate(text, lang, "hi")
        in ack_chars
    )


def sentence_split(text, lang, delim_pat="auto"):  ## New signature
    """split the text into sentences



    A rule-based sentence splitter for Indian languages written in

    Brahmi-derived scripts. The text is split at sentence delimiter

    boundaries. The delimiters can be configured by passing appropriate

    parameters.



    The sentence splitter can identify non-breaking phrases like

    single letter, common abbreviations/honorofics for some Indian

    languages.



    Args:

        text (str): text to split into sentence

        lang (str): ISO 639-2 language code

        delim_pat (str): regular expression to identify sentence delimiter characters. If set to 'auto', the delimiter pattern is chosen automatically based on the language and text.





    Returns:

        list: list of sentences identified from the input text

    """

    if lang == "ur":
        from indicnlp.urduhack.tokenization import sentence_tokenizer

        if len(text.split()) < 2:
            sentences = text.split()
        else:
            sentences = sentence_tokenizer(text)
        return sentences

    # print('Input: {}'.format(delim_pat))
    if delim_pat == "auto":
        if langinfo.is_danda_delim(lang):
            # in modern texts it is possible that period is used as delimeter
            # instead of DANDA. Hence, a check. Use danda delimiter pattern
            # only if text contains at least one danda
            if CONTAINS_DANDA.search(text) is None:
                delim_pat = DELIM_PAT_NO_DANDA
                # print('LANG has danda delim. TEXT_CONTAINS_DANDA: FALSE --> DELIM_PAT_NO_DANDA')
            else:
                delim_pat = DELIM_PAT_DANDA
                # print('LANG has danda delim. TEXT_CONTAINS_DANDA: TRUE --> DELIM_PAT_DANDA')
        else:
            delim_pat = DELIM_PAT_NO_DANDA
            # print('LANG has no danda delim --> DELIM_PAT_NO_DANDA')

    ## otherwise, assume the caller set the delimiter pattern

    ### Phase 1: break on sentence delimiters.
    cand_sentences = []
    begin = 0
    text = text.strip()
    for mo in delim_pat.finditer(text):
        p1 = mo.start()
        p2 = mo.end()

        ## NEW
        if p1 > 0 and text[p1 - 1].isnumeric():
            continue

        ## Prevents splitting on "." in URLs/emails in indic texts.
        if lang != "en":
            if is_latin_or_numeric(text[p1 - 1]):
                if p1 + 1 < len(text) and is_latin_or_numeric(text[p1 + 1]):
                    continue

        end = p1 + 1
        s = text[begin:end].strip()
        if len(s) > 0:
            cand_sentences.append(s)
        begin = p1 + 1

    s = text[begin:].strip()
    if len(s) > 0:
        cand_sentences.append(s)

    if not delim_pat.search("."):
        ## run phase 2 only if delimiter pattern contains period
        # print('No need to run phase2')
        return cand_sentences
    #     print(cand_sentences)
    #     print('====')

    #     return cand_sentences

    ### Phase 2: Address the fact that '.' may not always be a sentence delimiter
    ### Method: If there is a run of lines containing only a word (optionally) and '.',
    ### merge these lines as well one sentence preceding and succeeding this run of lines.
    final_sentences = []
    sen_buffer = ""
    bad_state = False

    for i, sentence in enumerate(cand_sentences):
        words = sentence.split(" ")
        # if len(words)<=2 and words[-1]=='.':
        if len(words) == 1 and sentence[-1] == ".":
            bad_state = True
            sen_buffer = sen_buffer + " " + sentence
        ## NEW condition
        elif sentence[-1] == "." and is_acronym_abbvr(words[-1][:-1], lang):
            if len(sen_buffer) > 0 and not bad_state:
                final_sentences.append(sen_buffer)
                sen_buffer = sentence
            else:
                sen_buffer = sen_buffer + " " + sentence
            bad_state = True
        elif bad_state:
            sen_buffer = sen_buffer + " " + sentence
            if len(sen_buffer) > 0:
                final_sentences.append(sen_buffer)
            sen_buffer = ""
            bad_state = False
        else:  ## good state
            if len(sen_buffer) > 0:
                final_sentences.append(sen_buffer)
            sen_buffer = sentence
            bad_state = False

    if len(sen_buffer) > 0:
        final_sentences.append(sen_buffer)

    for i in range(0, len(final_sentences)):
        final_sentences[i] = CONTAINS_MULTIPLE_SPACES.sub(
            " ", final_sentences[i].strip()
        )

    return final_sentences