czech: Add czech language support

Michal Bida · Michal Bida · commit 4613e4f27fec · 2024-09-30T13:18:32.000+02:00
- Based on the Slovak language that is very close to Czech one I have created initial support for Czech language sentence splitting
diff --git a/pysbd/lang/czech.py b/pysbd/lang/czech.py
@@ -0,0 +1,111 @@
+# -*- coding: utf-8 -*-
+import re
+from pysbd.abbreviation_replacer import AbbreviationReplacer
+from pysbd.between_punctuation import BetweenPunctuation
+from pysbd.lang.common import Common, Standard
+from pysbd.processor import Processor
+from pysbd.utils import Text
+from pysbd.punctuation_replacer import replace_punctuation
+from pysbd.lists_item_replacer import ListItemReplacer
+
+
+class Czech(Common, Standard):
+
+    iso_code = 'cz'
+
+    class ListItemReplacer(ListItemReplacer):
+
+        def add_line_break(self):
+            # We've found alphabetical lists are causing a lot of problems with abbreviations
+            # with multiple periods and spaces, such as 'Company name s. r. o.'. Disabling
+            # alphabetical list parsing seems like a reasonable tradeoff.
+
+            # self.format_alphabetical_lists()
+            self.format_roman_numeral_lists()
+            self.format_numbered_list_with_periods()
+            self.format_numbered_list_with_parens()
+            return self.text
+
+    class AbbreviationReplacer(AbbreviationReplacer):
+        SENTENCE_STARTERS = []
+
+        def replace_period_of_abbr(self, txt, abbr):
+            # This is a very simple version of the original function, which makes sure
+            # all of the periods in the abbreviation get replaced, not only the last one.
+            # In Czech language we use a lot of abbreviations like 'Company Name s. r. o.', so it
+            # is important to handle this properly.
+
+            abbr_new = abbr.replace(".", "∯") + "∯"
+            txt = txt.replace(abbr + ".", abbr_new)
+            return txt
+
+    class Abbreviation(Standard.Abbreviation):
+        ABBREVIATIONS = ['č', 'no', 'nr', 's. r. o', 'ing', 'p', 'a. d', 'o. k', 'pol. pr', 'a. s. a. p', 'p. n. l', 'red', 'o.k', 'a.d', 'm.o', 'pol.pr', 'a.s.a.p', 'p.n.l', 'pp', 'sl', 'corp', 'plgr', 'tz', 'rtg', 'o.c.p', 'o. c. p', 'c.k', 'c. k', 'n.a', 'n. a', 'a.m', 'a. m', 'vz', 'i.b', 'i. b', 'ú.p.v.o', 'ú. p. v. o', 'bros', 'rsdr', 'doc', 'tu', 'ods', 'n.w.a', 'n. w. a', 'nár', 'pedg', 'paeddr', 'rndr', 'naprk', 'napřk', 'a.g.p', 'a. g. p', 'prof', 'pr', 'př', 'a.v', 'a. v', 'por', 'mvdr', 'nešp', 'u.s', 'u. s', 'kt', 'vyd', 'e.t', 'e. t', 'al', 'll.m', 'll. m', 'o.f.i', 'o. f. i', 'mr', 'apod', 'súkr', 'střed', 's.e.g', 's. e. g', 'sr', 'tvz', 'ind', 'var', 'etc', 'atd', 'n.o', 'n. o', 's.a', 's. a', 'např', 'a.i.i', 'a. i. i', 'a.k.a', 'a. k. a', 'konkr', 'čsl', 'odd', 'ltd', 't.z', 't. z', 'o.z', 'o. z', 'obv', 'obr', 'pok', 'tel', 'št', 'skr', 'phdr', 'xx', 'š.p', 'š. p', 'ph.d', 'ph. d', 'm.n.m', 'm. n. m', 'zz', 'roz', 'atď.', 'ev', 'v.sp', 'v. sp', 'drsc', 'mudr', 't.č', 't. č', 'el', 'os', 'co', 'r.o', 'r. o', 'str', 'p.a', 'p. a', 'zdravot', 'prek', 'gen', 'viď', 'dr', 'cca', 'p.s', 'p. s', 'zák', 'slov', 'arm', 'inc', 'max', 'd.c', 'k.o', 'a. r. k', 'd. c', 'k. o', 'a. r. k', 'soc', 'bc', 'zs', 'akad', 'sz', 'pozn', 'tr', 'nám', 'kol', 'csc', 'ul', 'sp', 'o.i', 'jr', 'zb', 'sv', 'tj', 'čs', 'tzn', 'příp', 'iv', 'hl', 'st', 'pod', 'vi', 'tis', 'stor', 'rozh', 'mld', 'atď', 'mgr', 'a.s', 'a. s', 'phd', 'z.z', 'z. z', 'judr', 'ing', 'hod', 'vs', 'písm', 's.r.o', 'min', 'ml', 'iii', 't.j', 't. j', 'spol', 'mil', 'ii', 'napr', 'resp', 'tzv']
+        PREPOSITIVE_ABBREVIATIONS = ['st', 'p', 'dr', 'mudr', 'judr', 'ing', 'mgr', 'bc', 'drsc', 'doc', 'prof']
+        NUMBER_ABBREVIATIONS = ['č', 'no', 'nr']
+
+    class BetweenPunctuation(BetweenPunctuation):
+        # Rubular: https://rubular.com/r/rImWbaYFtHHtf4
+        BETWEEN_CZECH_DOUBLE_QUOTES_REGEX = r'„(?>[^“\\]+|\\{2}|\\.)*“'
+        BETWEEN_CZECH_DOUBLE_QUOTES_REGEX_2 = r'\„(?=(?P<tmp>[^“\\]+|\\{2}|\\.)*)(?P=tmp)\“'
+
+        def sub_punctuation_between_czech_double_quotes(self, txt):
+            return re.sub(self.BETWEEN_CZECH_DOUBLE_QUOTES_REGEX_2, replace_punctuation, txt)
+
+        def sub_punctuation_between_quotes_and_parens(self, txt):
+            txt = self.sub_punctuation_between_single_quotes(txt)
+            txt = self.sub_punctuation_between_single_quote_slanted(txt)
+            txt = self.sub_punctuation_between_double_quotes(txt)
+            txt = self.sub_punctuation_between_square_brackets(txt)
+            txt = self.sub_punctuation_between_parens(txt)
+            txt = self.sub_punctuation_between_quotes_arrow(txt)
+            txt = self.sub_punctuation_between_em_dashes(txt)
+            txt = self.sub_punctuation_between_quotes_slanted(txt)
+            txt = self.sub_punctuation_between_czech_double_quotes(txt)
+            return txt
+
+    class Processor(Processor):
+
+        def __init__(self, text, lang, char_span=False):
+            super().__init__(text, lang, char_span)
+
+        def process(self):
+            if not self.text:
+                return self.text
+            self.text = self.text.replace('\n', '\r')
+
+            # Here we use language specific ListItemReplacer:
+            li = self.lang.ListItemReplacer(self.text)
+            self.text = li.add_line_break()
+
+            self.replace_abbreviations()
+            self.replace_numbers()
+            self.replace_continuous_punctuation()
+            self.replace_periods_before_numeric_references()
+            self.text = Text(self.text).apply(
+                self.lang.Abbreviation.WithMultiplePeriodsAndEmailRule,
+                self.lang.GeoLocationRule, self.lang.FileFormatRule)
+            postprocessed_sents = self.split_into_segments()
+            return postprocessed_sents
+
+        def replace_numbers(self):
+            self.text = Text(self.text).apply(*self.lang.Numbers.All)
+            self.replace_period_in_czech_dates()
+            self.replace_period_in_ordinal_numerals()
+            self.replace_period_in_roman_numerals()
+            return self.text
+
+        def replace_period_in_ordinal_numerals(self):
+            # Rubular: https://rubular.com/r/0HkmvzMGTqgWs6
+            self.text = re.sub(r'(?<=\d)\.(?=\s*[a-z]+)', '∯', self.text)
+
+        def replace_period_in_roman_numerals(self):
+            # Rubular: https://rubular.com/r/XlzTIi7aBRThSl
+            self.text = re.sub(r'((\s+[VXI]+)|(^[VXI]+))(\.)(?=\s+)', r'\1∯', self.text, re.IGNORECASE)
+
+        def replace_period_in_czech_dates(self):
+            MONTHS = ['leden', 'únor', 'březen', 'duben', 'květen', 'červen', 'červenec', 'srpen', 'září', 'říjen', 'listopad', 'prosinec',
+                      'ledna', 'února', 'března', 'dubna', 'května', 'června', 'července', 'srpna', 'října', 'listopadu', 'prosince']
+            for month in MONTHS:
+                # Rubular: https://rubular.com/r/dGLZqsbjcdJvCd
+                self.text = re.sub(r'(?<=\d)\.(?=\s*{month})'.format(month=month), '∯', self.text)
diff --git a/pysbd/languages.py b/pysbd/languages.py
@@ -22,6 +22,7 @@
 from pysbd.lang.deutsch import Deutsch
 from pysbd.lang.kazakh import Kazakh
 from pysbd.lang.slovak import Slovak
+from pysbd.lang.czech import Czech
 
 LANGUAGE_CODES = {
     'en': English,
@@ -46,7 +47,8 @@
     'ja': Japanese,
     'de': Deutsch,
     'kk': Kazakh,
-    'sk': Slovak
+    'sk': Slovak,
+    'cz': Czech
 }
 
 
diff --git a/tests/conftest.py b/tests/conftest.py
@@ -146,3 +146,8 @@ def kk_default_fixture():
 def sk_default_fixture():
     sk_segmenter = pysbd.Segmenter(language="sk", clean=False, char_span=False)
     return sk_segmenter
+
+@pytest.fixture()
+def cz_default_fixture():
+    cz_segmenter = pysbd.Segmenter(language="cz", clean=False, char_span=False)
+    return cz_segmenter
diff --git a/tests/lang/test_czech.py b/tests/lang/test_czech.py
@@ -0,0 +1,23 @@
+# -*- coding: utf-8 -*-
+import pytest
+
+GOLDEN_CZ_RULES_TEST_CASES = [
+    ("Jde o majite firmy ABTrade s. r. o., kteří stojí i za dalšími společnostmi, např. XYZCorp a.s.",
+     ["Jde o majite firmy ABTrade s. r. o., kteří stojí i za dalšími společnostmi, např. XYZCorp a.s."]),
+    ("„Průzkumy beru na lehkou váhu. V podstatě mě to nezajímá,“ reagoval Zeman na průzkum agentury Focus.",
+     ["„Průzkumy beru na lehkou váhu. V podstatě mě to nezajímá,“ reagoval Zeman na průzkum agentury Focus."]),
+    ("Toto se mi podařilo až na 10. pokus, ale stálo to za to.",
+     ["Toto se mi podařilo až na 10. pokus, ale stálo to za to."]),
+    ("Jde o príslušníky XII. Pluku speciálního nasazení.",
+     ["Jde o príslušníky XII. Pluku speciálního nasazení."]),
+    ("Společnost byla založena 7. dubna 2020, na smlouvě však figuruje datum 20. březen 2020.",
+     ["Společnost byla založena 7. dubna 2020, na smlouvě však figuruje datum 20. březen 2020."]),
+]
+
+
+@pytest.mark.parametrize('text,expected_sents', GOLDEN_CZ_RULES_TEST_CASES)
+def test_pl_sbd(cz_default_fixture, text, expected_sents):
+    """Czech language SBD tests"""
+    segments = cz_default_fixture.segment(text)
+    segments = [s.strip() for s in segments]
+    assert segments == expected_sents
