Merge pull request #865 from pavaris-pm/update-pos-tag-transformers

Update `pos_tag_transformers` function

Author: wannaphong

pythainlp/tag/pos_tag.py

@@ -180,13 +180,36 @@ def pos_tag_sents(
 
 
 def pos_tag_transformers(
-    words: str, engine: str = "bert-base-th-cased-blackboard"
-):
+    sentence: str, 
+    engine: str = "bert",
+    corpus: str = "blackboard",
+)->List[List[Tuple[str, str]]]:
     """
-    "wangchanberta-ud-thai-pud-upos",
-    "mdeberta-v3-ud-thai-pud-upos",
-    "bert-base-th-cased-blackboard",
+    Marks sentences with part-of-speech (POS) tags.
+
+    :param str sentence: a list of lists of tokenized words
+    :param str engine:
+        * *bert* -  BERT: Bidirectional Encoder Representations from Transformers (default)
+        * *wangchanberta* - fine-tuned version of airesearch/wangchanberta-base-att-spm-uncased on pud corpus (support PUD cotpus only)
+        * *mdeberta* - mDeBERTa: Multilingual Decoding-enhanced BERT with disentangled attention (support PUD corpus only)
+    :param str corpus: the corpus that is used to create the language model for tagger
+        * *blackboard* - `blackboard treebank (support bert engine only) <https://bitbucket.org/kaamanita/blackboard-treebank/src/master/>`_
+        * *pud* - `Parallel Universal Dependencies (PUD)\
+            <https://github.com/UniversalDependencies/UD_Thai-PUD>`_ \
+            treebanks, natively use Universal POS tags (support wangchanberta and mdeberta engine)
+    :return: a list of lists of tuples (word, POS tag)
+    :rtype: list[list[tuple[str, str]]]
 
+    :Example:
+
+    Labels POS for given sentence::
+
+        from pythainlp.tag import pos_tag_transformers
+
+        sentences = "แมวทำอะไรตอนห้าโมงเช้า"
+        pos_tag_transformers(sentences, engine="bert", corpus='blackboard')
+        # output:
+        # [[('แมว', 'NOUN'), ('ทําอะไร', 'VERB'), ('ตอนห้าโมงเช้า', 'NOUN')]]
     """
 
     try:
@@ -196,28 +219,35 @@ def pos_tag_transformers(
         raise ImportError(
             "Not found transformers! Please install transformers by pip install transformers")
 
-    if not words:
+    if not sentence:
         return []
 
-    if engine == "wangchanberta-ud-thai-pud-upos":
-        model = AutoModelForTokenClassification.from_pretrained(
-            "Pavarissy/wangchanberta-ud-thai-pud-upos")
-        tokenizer = AutoTokenizer.from_pretrained("Pavarissy/wangchanberta-ud-thai-pud-upos")
-    elif engine == "mdeberta-v3-ud-thai-pud-upos":
-        model = AutoModelForTokenClassification.from_pretrained(
-            "Pavarissy/mdeberta-v3-ud-thai-pud-upos")
-        tokenizer = AutoTokenizer.from_pretrained("Pavarissy/mdeberta-v3-ud-thai-pud-upos")
-    elif engine == "bert-base-th-cased-blackboard":
-        model = AutoModelForTokenClassification.from_pretrained("lunarlist/pos_thai")
-        tokenizer = AutoTokenizer.from_pretrained("lunarlist/pos_thai")
+    _blackboard_support_engine = {
+        "bert" : "lunarlist/pos_thai",
+    }
+
+    _pud_support_engine = {
+        "wangchanberta" : "Pavarissy/wangchanberta-ud-thai-pud-upos",
+        "mdeberta" : "Pavarissy/mdeberta-v3-ud-thai-pud-upos",
+    }
+
+    if corpus == 'blackboard' and engine in _blackboard_support_engine.keys():
+        base_model = _blackboard_support_engine.get(engine)
+        model = AutoModelForTokenClassification.from_pretrained(base_model)
+        tokenizer = AutoTokenizer.from_pretrained(base_model)
+    elif corpus == 'pud' and engine in _pud_support_engine.keys():
+        base_model = _pud_support_engine.get(engine)
+        model = AutoModelForTokenClassification.from_pretrained(base_model)
+        tokenizer = AutoTokenizer.from_pretrained(base_model)
     else:
         raise ValueError(
-            "pos_tag_transformers not support {0} engine.".format(
-                engine
+            "pos_tag_transformers not support {0} engine or {1} corpus.".format(
+                engine, corpus
             )
         )
 
-    pipeline = TokenClassificationPipeline(model=model, tokenizer=tokenizer, grouped_entities=True)
+    pipeline = TokenClassificationPipeline(model=model, tokenizer=tokenizer, aggregation_strategy="simple")
 
-    outputs = pipeline(words)
-    return outputs
+    outputs = pipeline(sentence)
+    word_tags = [[(tag['word'], tag['entity_group']) for tag in outputs]]
+    return word_tags

tests/test_tag.py

@@ -367,10 +367,13 @@ def test_NNER_class(self):
 
     def test_pos_tag_transformers(self):
         self.assertIsNotNone(pos_tag_transformers(
-            words="แมวทำอะไรตอนห้าโมงเช้า", engine="bert-base-th-cased-blackboard"))
+            words="แมวทำอะไรตอนห้าโมงเช้า", engine="bert", corpus="blackboard"))
         self.assertIsNotNone(pos_tag_transformers(
-            words="แมวทำอะไรตอนห้าโมงเช้า", engine="mdeberta-v3-ud-thai-pud-upos"))
+            words="แมวทำอะไรตอนห้าโมงเช้า", engine="mdeberta", corpus="pud"))
         self.assertIsNotNone(pos_tag_transformers(
-            words="แมวทำอะไรตอนห้าโมงเช้า", engine="wangchanberta-ud-thai-pud-upos"))
+            words="แมวทำอะไรตอนห้าโมงเช้า", engine="wangchanberta", corpus="pud"))
         with self.assertRaises(ValueError):
-            pos_tag_transformers(words="แมวทำอะไรตอนห้าโมงเช้า", engine="non-existing-engine")
+            pos_tag_transformers(words="แมวทำอะไรตอนห้าโมงเช้า", engine="non-existing-engine")
+        with self.assertRaises(ValueError):
+            pos_tag_transformers(words="แมวทำอะไรตอนห้าโมงเช้า", engine="bert", 
+                                 corpus="non-existing corpus")