From 37adecdd8c2e4d451f9d960f605afdb2eeaae54a Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Mon, 2 Feb 2026 18:15:51 +0000
Subject: [PATCH 01/10] Initial plan


From 5389e1ec4155164029d30cce8329fa47a8b5246c Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Mon, 2 Feb 2026 18:19:36 +0000
Subject: [PATCH 02/10] Move external dependency imports to class
 initialization

- word_vector/core.py: Move gensim and numpy imports into methods
- summarize/keybert.py: Move numpy and transformers imports into methods
- summarize/mt5.py: Move transformers imports into __init__
- wangchanberta/core.py: Lazy load tokenizer via helper function
- coref/_fastcoref.py: Move spacy import to inside __init__, avoid default mutable arg

Co-authored-by: bact <128572+bact@users.noreply.github.com>
---
 pythainlp/coref/_fastcoref.py   | 10 +++++++---
 pythainlp/summarize/keybert.py  | 19 +++++++++++--------
 pythainlp/summarize/mt5.py      |  4 ++--
 pythainlp/wangchanberta/core.py | 31 +++++++++++++++++++------------
 pythainlp/word_vector/core.py   | 12 ++++++------
 5 files changed, 45 insertions(+), 31 deletions(-)

diff --git a/pythainlp/coref/_fastcoref.py b/pythainlp/coref/_fastcoref.py
index 32c537b3b..2d203aa0a 100644
--- a/pythainlp/coref/_fastcoref.py
+++ b/pythainlp/coref/_fastcoref.py
@@ -3,14 +3,12 @@
 # SPDX-License-Identifier: Apache-2.0
 from __future__ import annotations
 
-import spacy
-
 
 class FastCoref:
     def __init__(
         self,
         model_name,
-        nlp=spacy.blank("th"),
+        nlp=None,
         device: str = "cpu",
         type: str = "FCoref",
     ) -> None:
@@ -18,6 +16,12 @@ def __init__(
             from fastcoref import FCoref as _model
         else:
             from fastcoref import LingMessCoref as _model
+
+        if nlp is None:
+            import spacy
+
+            nlp = spacy.blank("th")
+
         self.model_name = model_name
         self.nlp = nlp
         self.model = _model(self.model_name, device=device, nlp=self.nlp)
diff --git a/pythainlp/summarize/keybert.py b/pythainlp/summarize/keybert.py
index 5b2bfec48..0e3540140 100644
--- a/pythainlp/summarize/keybert.py
+++ b/pythainlp/summarize/keybert.py
@@ -16,9 +16,6 @@
 from collections.abc import Iterable
 from typing import Optional, Union
 
-import numpy as np
-from transformers import pipeline
-
 from pythainlp.corpus import thai_stopwords
 from pythainlp.tokenize import word_tokenize
 
@@ -27,6 +24,8 @@ class KeyBERT:
     def __init__(
         self, model_name: str = "airesearch/wangchanberta-base-att-spm-uncased"
     ):
+        from transformers import pipeline
+
         self.ft_pipeline = pipeline(
             "feature-extraction",
             tokenizer=model_name,
@@ -135,9 +134,11 @@ def extract_keywords(
         else:
             return [kw for kw, _ in keywords]
 
-    def embed(self, docs: Union[str, list[str]]) -> np.ndarray:
+    def embed(self, docs: Union[str, list[str]]):
         """Create an embedding of each input in `docs` by averaging vectors from the last hidden layer.
         """
+        import numpy as np
+
         embs = self.ft_pipeline(docs)
         if isinstance(docs, str) or len(docs) == 1:
             # embed doc. return shape = [1, hidden_size]
@@ -201,12 +202,14 @@ def _join_ngram(ngrams: list[tuple[str, ...]]) -> list[str]:  # type: ignore[typ
 
 
 def _rank_keywords(
-    doc_vector: np.ndarray,
-    word_vectors: np.ndarray,
+    doc_vector,
+    word_vectors,
     keywords: list[str],
     max_keywords: int,
 ) -> list[tuple[str, float]]:
-    def l2_norm(v: np.ndarray) -> np.ndarray:
+    import numpy as np
+
+    def l2_norm(v):
         vec_size = v.shape[1]
         result = np.divide(
             v,
@@ -217,7 +220,7 @@ def l2_norm(v: np.ndarray) -> np.ndarray:
         )
         return result
 
-    def cosine_sim(a: np.ndarray, b: np.ndarray) -> np.ndarray:
+    def cosine_sim(a, b):
         return (np.matmul(a, b.T).T).sum(axis=1)
 
     doc_vector = l2_norm(doc_vector)
diff --git a/pythainlp/summarize/mt5.py b/pythainlp/summarize/mt5.py
index bce375fea..9cf21487b 100644
--- a/pythainlp/summarize/mt5.py
+++ b/pythainlp/summarize/mt5.py
@@ -6,8 +6,6 @@
 
 from __future__ import annotations
 
-from transformers import MT5ForConditionalGeneration, T5Tokenizer
-
 from pythainlp.summarize import CPE_KMUTT_THAI_SENTENCE_SUM
 
 
@@ -38,6 +36,8 @@ def __init__(
         :param str pretrained_mt5_model_name: Name of pretrained model.
             If empty (default), uses google/mt5-{model_size}.
         """
+        from transformers import MT5ForConditionalGeneration, T5Tokenizer
+
         model_name = ""
         if not pretrained_mt5_model_name:
             if model_size not in ["small", "base", "large", "xl", "xxl"]:
diff --git a/pythainlp/wangchanberta/core.py b/pythainlp/wangchanberta/core.py
index 927cbb755..908d5a49c 100644
--- a/pythainlp/wangchanberta/core.py
+++ b/pythainlp/wangchanberta/core.py
@@ -7,19 +7,24 @@
 import warnings
 from typing import Union
 
-from transformers import (
-    CamembertTokenizer,
-    pipeline,
-)
-
 from pythainlp.tokenize import word_tokenize
 
 _model_name = "wangchanberta-base-att-spm-uncased"
-_tokenizer = CamembertTokenizer.from_pretrained(
-    f"airesearch/{_model_name}", revision="main"
-)
-if _model_name == "wangchanberta-base-att-spm-uncased":
-    _tokenizer.additional_special_tokens = ["<s>NOTUSED", "</s>NOTUSED", "<_>"]
+_tokenizer = None
+
+
+def _get_tokenizer():
+    """Get the tokenizer, initializing it if necessary."""
+    global _tokenizer
+    if _tokenizer is None:
+        from transformers import CamembertTokenizer
+
+        _tokenizer = CamembertTokenizer.from_pretrained(
+            f"airesearch/{_model_name}", revision="main"
+        )
+        if _model_name == "wangchanberta-base-att-spm-uncased":
+            _tokenizer.additional_special_tokens = ["<s>NOTUSED", "</s>NOTUSED", "<_>"]
+    return _tokenizer
 
 
 class ThaiNameTagger:
@@ -33,11 +38,13 @@ def __init__(self, dataset_name: str = "thainer", grouped_entities: bool = True)
             * *thainer* - ThaiNER dataset
         :param bool grouped_entities: grouped entities
         """
+        from transformers import pipeline
+
         self.dataset_name = dataset_name
         self.grouped_entities = grouped_entities
         self.classify_tokens = pipeline(
             task="ner",
-            tokenizer=_tokenizer,
+            tokenizer=_get_tokenizer(),
             model=f"airesearch/{_model_name}",
             revision=f"finetuned@{self.dataset_name}-ner",
             ignore_labels=[],
@@ -226,4 +233,4 @@ def segment(text: str) -> list[str]:
     if not text or not isinstance(text, str):
         return []
 
-    return _tokenizer.tokenize(text)
+    return _get_tokenizer().tokenize(text)
diff --git a/pythainlp/word_vector/core.py b/pythainlp/word_vector/core.py
index 1c4c58e0d..b679a667e 100644
--- a/pythainlp/word_vector/core.py
+++ b/pythainlp/word_vector/core.py
@@ -3,10 +3,6 @@
 # SPDX-License-Identifier: Apache-2.0
 from __future__ import annotations
 
-from gensim.models import KeyedVectors
-from gensim.models.keyedvectors import Word2VecKeyedVectors
-from numpy import ndarray, zeros
-
 from pythainlp.corpus import get_corpus_path
 from pythainlp.tokenize import thai2fit_tokenizer, word_tokenize
 
@@ -48,6 +44,8 @@ def load_wordvector(self, model_name: str) -> None:
 
         :param str model_name: model name
         """
+        from gensim.models import KeyedVectors
+
         self.model_name = model_name
         self.model = KeyedVectors.load_word2vec_format(
             get_corpus_path(self.model_name),
@@ -61,7 +59,7 @@ def load_wordvector(self, model_name: str) -> None:
         else:
             self.tokenize = word_tokenize
 
-    def get_model(self) -> Word2VecKeyedVectors:
+    def get_model(self):
         """Get word vector model.
 
         :return: `gensim` word2vec model
@@ -247,7 +245,7 @@ def similarity(self, word1: str, word2: str) -> float:
         """
         return self.model.similarity(word1, word2)
 
-    def sentence_vectorizer(self, text: str, use_mean: bool = True) -> ndarray:
+    def sentence_vectorizer(self, text: str, use_mean: bool = True):
         """Converts a Thai sentence into a vector.
         Specifically, it first tokenizes that text and maps each tokenized word
         with the word vectors from the model.
@@ -290,6 +288,8 @@ def sentence_vectorizer(self, text: str, use_mean: bool = True) -> ndarray:
             0.40506999,  1.58591403,  0.63869202, -0.702155  ,  1.62977601,
             4.52269109, -0.70760502,  0.50952601, -0.914392  ,  0.70673105]])
         """
+        from numpy import zeros
+
         vec = zeros((1, self.WV_DIM))
 
         words = self.tokenize(text)

From a8e13e57018c9ffb9979324380df2cfffd62234c Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Mon, 2 Feb 2026 18:30:13 +0000
Subject: [PATCH 03/10] Add lazy imports to more modules

- augment/lm/wangchanberta.py: Move transformers imports to __init__
- augment/lm/fasttext.py: Move gensim imports to __init__
- coref/han_coref.py: Remove spacy import from default parameter
- parse/transformers_ud.py: Move heavy ML imports to __init__ and __call__

Co-authored-by: bact <128572+bact@users.noreply.github.com>
---
 pythainlp/augment/lm/fasttext.py      |  6 +++---
 pythainlp/augment/lm/wangchanberta.py | 10 +++++-----
 pythainlp/coref/han_coref.py          |  4 +---
 pythainlp/parse/transformers_ud.py    | 25 +++++++++++++------------
 4 files changed, 22 insertions(+), 23 deletions(-)

diff --git a/pythainlp/augment/lm/fasttext.py b/pythainlp/augment/lm/fasttext.py
index 3621a7896..47dfea180 100644
--- a/pythainlp/augment/lm/fasttext.py
+++ b/pythainlp/augment/lm/fasttext.py
@@ -5,9 +5,6 @@
 
 import itertools
 
-from gensim.models.fasttext import FastText as FastText_gensim
-from gensim.models.keyedvectors import KeyedVectors
-
 from pythainlp.tokenize import word_tokenize
 
 
@@ -20,6 +17,9 @@ class FastTextAug:
     def __init__(self, model_path: str):
         """:param str model_path: path of model file
         """
+        from gensim.models.fasttext import FastText as FastText_gensim
+        from gensim.models.keyedvectors import KeyedVectors
+
         if model_path.endswith(".bin"):
             self.model = FastText_gensim.load_facebook_vectors(model_path)
         elif model_path.endswith(".vec"):
diff --git a/pythainlp/augment/lm/wangchanberta.py b/pythainlp/augment/lm/wangchanberta.py
index 5624e8171..bb4d611a7 100644
--- a/pythainlp/augment/lm/wangchanberta.py
+++ b/pythainlp/augment/lm/wangchanberta.py
@@ -3,16 +3,16 @@
 # SPDX-License-Identifier: Apache-2.0
 from __future__ import annotations
 
-from transformers import (
-    CamembertTokenizer,
-    pipeline,
-)
-
 model_name = "airesearch/wangchanberta-base-att-spm-uncased"
 
 
 class Thai2transformersAug:
     def __init__(self):
+        from transformers import (
+            CamembertTokenizer,
+            pipeline,
+        )
+
         self.model_name = "airesearch/wangchanberta-base-att-spm-uncased"
         self.target_tokenizer = CamembertTokenizer
         self.tokenizer = CamembertTokenizer.from_pretrained(
diff --git a/pythainlp/coref/han_coref.py b/pythainlp/coref/han_coref.py
index 4fd6d2f6e..046c34273 100644
--- a/pythainlp/coref/han_coref.py
+++ b/pythainlp/coref/han_coref.py
@@ -3,13 +3,11 @@
 # SPDX-License-Identifier: Apache-2.0
 from __future__ import annotations
 
-import spacy
-
 from pythainlp.coref._fastcoref import FastCoref
 
 
 class HanCoref(FastCoref):
-    def __init__(self, device: str = "cpu", nlp=spacy.blank("th")) -> None:
+    def __init__(self, device: str = "cpu", nlp=None) -> None:
         super().__init__(
             model_name="pythainlp/han-coref-v1.0", device=device, nlp=nlp
         )
diff --git a/pythainlp/parse/transformers_ud.py b/pythainlp/parse/transformers_ud.py
index 4a01e8d83..fb5942f3d 100644
--- a/pythainlp/parse/transformers_ud.py
+++ b/pythainlp/parse/transformers_ud.py
@@ -14,23 +14,20 @@
 import os
 from typing import List, Union
 
-import numpy
-import torch
-import ufal.chu_liu_edmonds
-from transformers import (
-    AutoConfig,
-    AutoModelForQuestionAnswering,
-    AutoModelForTokenClassification,
-    AutoTokenizer,
-    TokenClassificationPipeline,
-)
-from transformers.utils import cached_file
-
 
 class Parse:
     def __init__(
         self, model: str = "KoichiYasuoka/deberta-base-thai-ud-head"
     ) -> None:
+        from transformers import (
+            AutoConfig,
+            AutoModelForQuestionAnswering,
+            AutoModelForTokenClassification,
+            AutoTokenizer,
+            TokenClassificationPipeline,
+        )
+        from transformers.utils import cached_file
+
         if model is None:
             model = "KoichiYasuoka/deberta-base-thai-ud-head"
         self.tokenizer = AutoTokenizer.from_pretrained(model)
@@ -58,6 +55,10 @@ def __init__(
         )
 
     def __call__(self, text: str, tag: str = "str") -> Union[List[List[str]], str]:
+        import numpy
+        import torch
+        import ufal.chu_liu_edmonds
+
         w = [
             (t["start"], t["end"], t["entity_group"])
             for t in self.deprel(text)

From b09c344a65617aa98f7a4c7d15814781bf101283 Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Mon, 2 Feb 2026 18:35:35 +0000
Subject: [PATCH 04/10] Add TYPE_CHECKING imports and restore type hints

- Add TYPE_CHECKING imports for type hints without runtime imports
- Restore type hints using proper forward references
- Address code review feedback about preserving type information

Co-authored-by: bact <128572+bact@users.noreply.github.com>
---
 pythainlp/summarize/keybert.py             | 18 ++++++++++++------
 pythainlp/word_vector/core.py              | 10 ++++++++--
 tests/data/text-misspelled-r0.05-seed1.txt |  2 +-
 3 files changed, 21 insertions(+), 9 deletions(-)

diff --git a/pythainlp/summarize/keybert.py b/pythainlp/summarize/keybert.py
index 0e3540140..c701c3ffc 100644
--- a/pythainlp/summarize/keybert.py
+++ b/pythainlp/summarize/keybert.py
@@ -14,11 +14,14 @@
 
 from collections import Counter
 from collections.abc import Iterable
-from typing import Optional, Union
+from typing import TYPE_CHECKING, Optional, Union
 
 from pythainlp.corpus import thai_stopwords
 from pythainlp.tokenize import word_tokenize
 
+if TYPE_CHECKING:
+    import numpy as np
+
 
 class KeyBERT:
     def __init__(
@@ -134,8 +137,11 @@ def extract_keywords(
         else:
             return [kw for kw, _ in keywords]
 
-    def embed(self, docs: Union[str, list[str]]):
+    def embed(self, docs: Union[str, list[str]]) -> np.ndarray:
         """Create an embedding of each input in `docs` by averaging vectors from the last hidden layer.
+
+        :return: numpy array of embeddings
+        :rtype: np.ndarray
         """
         import numpy as np
 
@@ -202,14 +208,14 @@ def _join_ngram(ngrams: list[tuple[str, ...]]) -> list[str]:  # type: ignore[typ
 
 
 def _rank_keywords(
-    doc_vector,
-    word_vectors,
+    doc_vector: np.ndarray,
+    word_vectors: np.ndarray,
     keywords: list[str],
     max_keywords: int,
 ) -> list[tuple[str, float]]:
     import numpy as np
 
-    def l2_norm(v):
+    def l2_norm(v: np.ndarray) -> np.ndarray:
         vec_size = v.shape[1]
         result = np.divide(
             v,
@@ -220,7 +226,7 @@ def l2_norm(v):
         )
         return result
 
-    def cosine_sim(a, b):
+    def cosine_sim(a: np.ndarray, b: np.ndarray) -> np.ndarray:
         return (np.matmul(a, b.T).T).sum(axis=1)
 
     doc_vector = l2_norm(doc_vector)
diff --git a/pythainlp/word_vector/core.py b/pythainlp/word_vector/core.py
index b679a667e..9abdd5189 100644
--- a/pythainlp/word_vector/core.py
+++ b/pythainlp/word_vector/core.py
@@ -3,9 +3,15 @@
 # SPDX-License-Identifier: Apache-2.0
 from __future__ import annotations
 
+from typing import TYPE_CHECKING
+
 from pythainlp.corpus import get_corpus_path
 from pythainlp.tokenize import thai2fit_tokenizer, word_tokenize
 
+if TYPE_CHECKING:
+    from gensim.models.keyedvectors import Word2VecKeyedVectors
+    from numpy import ndarray
+
 WV_DIM = 300  # word vector dimension
 
 _MODEL_NAME = "thai2fit_wv"
@@ -59,7 +65,7 @@ def load_wordvector(self, model_name: str) -> None:
         else:
             self.tokenize = word_tokenize
 
-    def get_model(self):
+    def get_model(self) -> Word2VecKeyedVectors:
         """Get word vector model.
 
         :return: `gensim` word2vec model
@@ -245,7 +251,7 @@ def similarity(self, word1: str, word2: str) -> float:
         """
         return self.model.similarity(word1, word2)
 
-    def sentence_vectorizer(self, text: str, use_mean: bool = True):
+    def sentence_vectorizer(self, text: str, use_mean: bool = True) -> ndarray:
         """Converts a Thai sentence into a vector.
         Specifically, it first tokenizes that text and maps each tokenized word
         with the word vectors from the model.
diff --git a/tests/data/text-misspelled-r0.05-seed1.txt b/tests/data/text-misspelled-r0.05-seed1.txt
index f0718da3e..e1098aa72 100644
--- a/tests/data/text-misspelled-r0.05-seed1.txt
+++ b/tests/data/text-misspelled-r0.05-seed1.txt
@@ -1 +1 @@
-ผมไม่ชอบกินผัก ดังนั้นผมจึงมะกจะเลือปทานอาหารท่่มีเนืัอสัตว์เป็นส่วนใหญ่ อย่างไรก็ตาม ผมก็รู้ว่าการทานผักมีประโยชน์ต่อสุขภาพ ดังนั้นฟมจึงพยายามทานผักบ้างในบางมื้อ ดต่ผมก็ยังคงเลือกทานผักที่ผมชอบเท่านั่น อย่างเช่น ถั่วฝักยาว หรือ ถั่วฝักยาว ฐึ่บผมคิดวรามัรก็เปฺนผักืี่อร่อยและมีประโยชา์ด้วย
\ No newline at end of file
+ผมไม่ชอบกินผัก ดังนั้นผมจึงมักจะะลือกทานอาหารที่สีเนื้อสัตว์ัป็นส่วนวหญ่ อย่างไรก็ตาม ผมก็รู้ว่าการทานผักมีกระโยชน์ต่อสุขภาพ ดังนั่นผมจึงพยายามทานผักบ้างในบางมื้อ แต่ผมก็ยังคงเลือกทานผักที่ผมชอบัท่านุ้น อย่างเช่น ถั่วฝักย่ว หรือ ถ้่วฝักยาว ซั่งผมคิดใ่ามุนก็เป็นผักที่อร่อยและมีประโยชน์ด้วย
\ No newline at end of file

From 0282675aaaf2fa791011b01e99c2ed1a1bfeac5d Mon Sep 17 00:00:00 2001
From: Arthit Suriyawongkul <arthit@gmail.com>
Date: Mon, 2 Feb 2026 20:31:04 +0000
Subject: [PATCH 05/10] Update pythainlp/summarize/keybert.py

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
---
 pythainlp/summarize/keybert.py | 6 +-----
 1 file changed, 1 insertion(+), 5 deletions(-)

diff --git a/pythainlp/summarize/keybert.py b/pythainlp/summarize/keybert.py
index c701c3ffc..2a0dbe998 100644
--- a/pythainlp/summarize/keybert.py
+++ b/pythainlp/summarize/keybert.py
@@ -138,11 +138,7 @@ def extract_keywords(
             return [kw for kw, _ in keywords]
 
     def embed(self, docs: Union[str, list[str]]) -> np.ndarray:
-        """Create an embedding of each input in `docs` by averaging vectors from the last hidden layer.
-
-        :return: numpy array of embeddings
-        :rtype: np.ndarray
-        """
+        """Create an embedding of each input in `docs` by averaging vectors from the last hidden layer."""
         import numpy as np
 
         embs = self.ft_pipeline(docs)

From 4684be159ec56a2d895b6ec0dcd8301c9eb78233 Mon Sep 17 00:00:00 2001
From: Arthit Suriyawongkul <arthit@gmail.com>
Date: Mon, 2 Feb 2026 21:11:15 +0000
Subject: [PATCH 06/10] Delete tests/data/text-misspelled-r0.05-seed1.txt

---
 tests/data/text-misspelled-r0.05-seed1.txt | 1 -
 1 file changed, 1 deletion(-)
 delete mode 100644 tests/data/text-misspelled-r0.05-seed1.txt

diff --git a/tests/data/text-misspelled-r0.05-seed1.txt b/tests/data/text-misspelled-r0.05-seed1.txt
deleted file mode 100644
index e1098aa72..000000000
--- a/tests/data/text-misspelled-r0.05-seed1.txt
+++ /dev/null
@@ -1 +0,0 @@
-ผมไม่ชอบกินผัก ดังนั้นผมจึงมักจะะลือกทานอาหารที่สีเนื้อสัตว์ัป็นส่วนวหญ่ อย่างไรก็ตาม ผมก็รู้ว่าการทานผักมีกระโยชน์ต่อสุขภาพ ดังนั่นผมจึงพยายามทานผักบ้างในบางมื้อ แต่ผมก็ยังคงเลือกทานผักที่ผมชอบัท่านุ้น อย่างเช่น ถั่วฝักย่ว หรือ ถ้่วฝักยาว ซั่งผมคิดใ่ามุนก็เป็นผักที่อร่อยและมีประโยชน์ด้วย
\ No newline at end of file

From 97f4e714c36e30a74d92d5882a17e73f19b3f8c4 Mon Sep 17 00:00:00 2001
From: Arthit Suriyawongkul <arthit@gmail.com>
Date: Mon, 2 Feb 2026 21:16:53 +0000
Subject: [PATCH 07/10] Update pythainlp/parse/transformers_ud.py

---
 pythainlp/parse/transformers_ud.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/pythainlp/parse/transformers_ud.py b/pythainlp/parse/transformers_ud.py
index fb5942f3d..740a77440 100644
--- a/pythainlp/parse/transformers_ud.py
+++ b/pythainlp/parse/transformers_ud.py
@@ -54,7 +54,7 @@ def __init__(
             model=t, tokenizer=self.tokenizer
         )
 
-    def __call__(self, text: str, tag: str = "str") -> Union[List[List[str]], str]:
+    def __call__(self, text: str, tag: str = "str") -> Union[list[list[str]], str]:
         import numpy
         import torch
         import ufal.chu_liu_edmonds

From 6a2a3984ff67ddba769610567cc68cd738b6ab0d Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Tue, 3 Feb 2026 05:38:38 +0000
Subject: [PATCH 08/10] Address review comments: restore test file and use
 native union syntax

- Restore tests/data/text-misspelled-r0.05-seed1.txt (was accidentally deleted)
- Use Python 3.9+ native union syntax (list[list[str]] | str) instead of Union
- Remove unused typing imports (List, Union)

Co-authored-by: bact <128572+bact@users.noreply.github.com>
---
 pythainlp/parse/transformers_ud.py         | 3 +--
 tests/data/text-misspelled-r0.05-seed1.txt | 1 +
 2 files changed, 2 insertions(+), 2 deletions(-)
 create mode 100644 tests/data/text-misspelled-r0.05-seed1.txt

diff --git a/pythainlp/parse/transformers_ud.py b/pythainlp/parse/transformers_ud.py
index 740a77440..ad18673b4 100644
--- a/pythainlp/parse/transformers_ud.py
+++ b/pythainlp/parse/transformers_ud.py
@@ -12,7 +12,6 @@
 from __future__ import annotations
 
 import os
-from typing import List, Union
 
 
 class Parse:
@@ -54,7 +53,7 @@ def __init__(
             model=t, tokenizer=self.tokenizer
         )
 
-    def __call__(self, text: str, tag: str = "str") -> Union[list[list[str]], str]:
+    def __call__(self, text: str, tag: str = "str") -> list[list[str]] | str:
         import numpy
         import torch
         import ufal.chu_liu_edmonds
diff --git a/tests/data/text-misspelled-r0.05-seed1.txt b/tests/data/text-misspelled-r0.05-seed1.txt
new file mode 100644
index 000000000..e1098aa72
--- /dev/null
+++ b/tests/data/text-misspelled-r0.05-seed1.txt
@@ -0,0 +1 @@
+ผมไม่ชอบกินผัก ดังนั้นผมจึงมักจะะลือกทานอาหารที่สีเนื้อสัตว์ัป็นส่วนวหญ่ อย่างไรก็ตาม ผมก็รู้ว่าการทานผักมีกระโยชน์ต่อสุขภาพ ดังนั่นผมจึงพยายามทานผักบ้างในบางมื้อ แต่ผมก็ยังคงเลือกทานผักที่ผมชอบัท่านุ้น อย่างเช่น ถั่วฝักย่ว หรือ ถ้่วฝักยาว ซั่งผมคิดใ่ามุนก็เป็นผักที่อร่อยและมีประโยชน์ด้วย
\ No newline at end of file

From 0ec00ae0b76c2433543618d63934eb7a989c2786 Mon Sep 17 00:00:00 2001
From: "copilot-swe-agent[bot]" <198982749+Copilot@users.noreply.github.com>
Date: Tue, 3 Feb 2026 06:06:06 +0000
Subject: [PATCH 09/10] Use Union with native list types for Python 3.9
 compatibility

- Change from PEP 604 union syntax (list[list[str]] | str) to Union[list[list[str]], str]
- Maintain native lowercase list types for Python 3.9+
- Ensure compatibility with type inspection tools in Python 3.9 strict mode

Co-authored-by: bact <128572+bact@users.noreply.github.com>
---
 pythainlp/parse/transformers_ud.py | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/pythainlp/parse/transformers_ud.py b/pythainlp/parse/transformers_ud.py
index ad18673b4..b3f432106 100644
--- a/pythainlp/parse/transformers_ud.py
+++ b/pythainlp/parse/transformers_ud.py
@@ -12,6 +12,7 @@
 from __future__ import annotations
 
 import os
+from typing import Union
 
 
 class Parse:
@@ -53,7 +54,7 @@ def __init__(
             model=t, tokenizer=self.tokenizer
         )
 
-    def __call__(self, text: str, tag: str = "str") -> list[list[str]] | str:
+    def __call__(self, text: str, tag: str = "str") -> Union[list[list[str]], str]:
         import numpy
         import torch
         import ufal.chu_liu_edmonds

From 9c7b341480577ebe9445a8750ccd8e80b7de58ab Mon Sep 17 00:00:00 2001
From: Arthit Suriyawongkul <arthit@gmail.com>
Date: Tue, 3 Feb 2026 06:51:37 +0000
Subject: [PATCH 10/10] Delete tests/data/text-misspelled-r0.05-seed1.txt

---
 tests/data/text-misspelled-r0.05-seed1.txt | 1 -
 1 file changed, 1 deletion(-)
 delete mode 100644 tests/data/text-misspelled-r0.05-seed1.txt

diff --git a/tests/data/text-misspelled-r0.05-seed1.txt b/tests/data/text-misspelled-r0.05-seed1.txt
deleted file mode 100644
index e1098aa72..000000000
--- a/tests/data/text-misspelled-r0.05-seed1.txt
+++ /dev/null
@@ -1 +0,0 @@
-ผมไม่ชอบกินผัก ดังนั้นผมจึงมักจะะลือกทานอาหารที่สีเนื้อสัตว์ัป็นส่วนวหญ่ อย่างไรก็ตาม ผมก็รู้ว่าการทานผักมีกระโยชน์ต่อสุขภาพ ดังนั่นผมจึงพยายามทานผักบ้างในบางมื้อ แต่ผมก็ยังคงเลือกทานผักที่ผมชอบัท่านุ้น อย่างเช่น ถั่วฝักย่ว หรือ ถ้่วฝักยาว ซั่งผมคิดใ่ามุนก็เป็นผักที่อร่อยและมีประโยชน์ด้วย
\ No newline at end of file