add thai ICU corpus

pythainlp/corpus/__init__.py

@@ -26,6 +26,7 @@
     "thai_dict",
     "thai_family_names",
     "thai_female_names",
+    "thai_icu",
     "thai_male_names",
     "thai_negations",
     "thai_orst_words",
@@ -110,5 +111,6 @@ def corpus_db_path() -> str:
     thai_words,
     thai_wsd_dict,
 )
+from pythainlp.corpus.thai_icu import thai_icu
 from pythainlp.corpus.volubilis import volubilis
 from pythainlp.corpus.wikipedia_titles import wikipedia_titles

pythainlp/corpus/thai_icu.py

@@ -0,0 +1,29 @@
+# -*- coding: utf-8 -*-
+# SPDX-FileCopyrightText: Copyright 2016-2023 PyThaiNLP Project
+# SPDX-License-Identifier: Apache-2.0
+"""
+Provides an optional word list from International Components for Unicode (ICU) dictionary.
+"""
+from typing import FrozenSet
+
+from pythainlp.corpus.common import get_corpus
+
+_THAI_ICU = None
+_THAI_ICU_FILENAME = "thai_icu.txt"
+
+def thai_icu() -> FrozenSet[str]:
+    """
+    Return a frozenset of words from the International Components for Unicode (ICU) dictionary.
+
+    The data is at pythainlp/corpus/thai_icu.txt
+    The word list has beed prepared by the code at:
+    https://github.com/unicode-org/icu/blob/main/icu4c/source/data/brkitr/dictionaries/thaidict.txt
+
+    :return: :class:`frozenset` containing words in the International Components for Unicode (ICU) dictionary.
+    :rtype: :class:`frozenset`
+    """
+    global _THAI_ICU
+    if not _THAI_ICU:
+        _THAI_ICU = get_corpus(_THAI_ICU_FILENAME)
+
+    return _THAI_ICU