Merge pull request #927 from ayaan-qadri/dev

Added list of string support to sent_tokenize

pythainlp/tokenize/core.py

@@ -6,6 +6,7 @@
 """
 import re
 from typing import Iterable, List, Union
+import copy
 
 from pythainlp.tokenize import (
     DEFAULT_SENT_TOKENIZE_ENGINE,
@@ -198,7 +199,7 @@ def word_tokenize(
 
         word_tokenize(text, engine="newmm", keep_whitespace=False)
         # output: ['วรรณกรรม', 'ภาพวาด', 'และ', 'การแสดง', 'งิ้ว']
-        
+
     Join broken formatted numeric (e.g. time, decimals, IP addresses)::
 
         text = "เงิน1,234บาท19:32น 127.0.0.1"
@@ -322,17 +323,50 @@ def word_tokenize(
     return segments
 
 
+def indices_words(words):
+    indices = []
+    start_index = 0
+    for word in words:
+        end_index = start_index + len(word) - 1
+        indices.append((start_index, end_index))
+        start_index += len(word)
+
+    return indices
+
+
+def map_indices_to_words(index_list, sentences):
+    result = []
+    c = copy.copy(index_list)
+    n_sum = 0
+    for sentence in sentences:
+        words = sentence
+        sentence_result = []
+        n = 0
+        for start, end in c:
+            if start > n_sum + len(words) - 1:
+                break
+            else:
+                word = sentence[start - n_sum:end + 1 - n_sum]
+                sentence_result.append(word)
+                n += 1
+
+        result.append(sentence_result)
+        n_sum += len(words)
+        for _ in range(n):
+            del c[0]
+    return result
+
 def sent_tokenize(
-    text: str,
+    text: Union[str, List[str]],
     engine: str = DEFAULT_SENT_TOKENIZE_ENGINE,
     keep_whitespace: bool = True,
 ) -> List[str]:
     """
     Sentence tokenizer.
 
-    Tokenizes running text into "sentences"
+    Tokenizes running text into "sentences". Supports both string and list of strings.
 
-    :param str text: the text to be tokenized
+    :param text: the text (string) or list of words (list of strings) to be tokenized
     :param str engine: choose among *'crfcut'*, *'whitespace'*, \
     *'whitespace+newline'*
     :return: list of split sentences
@@ -394,38 +428,84 @@ def sent_tokenize(
         'และเขาได้รับมอบหมายให้ประจำในระดับภูมิภาค']
     """
 
-    if not text or not isinstance(text, str):
+    if not text or not isinstance(text, (str, list)):
         return []
 
+    is_list_input = isinstance(text, list)
+
+    if is_list_input:
+
+        try:
+            original_text = "".join(text)
+        except ValueError:
+            return []
+
+    else:
+        original_text = text
+
     segments = []
 
     if engine == "crfcut":
         from pythainlp.tokenize.crfcut import segment
 
-        segments = segment(text)
+        segments = segment(original_text)
+
+        if is_list_input:
+            word_indices = indices_words(text)
+            result = map_indices_to_words(word_indices, [original_text])
+            return result
     elif engine == "whitespace":
-        segments = re.split(r" +", text, flags=re.U)
+        segments = re.split(r" +", original_text, flags=re.U)
+        if is_list_input:
+            result = []
+            _temp = []
+            for i, w in enumerate(text):
+                if re.findall(r" ", w) != [] and re.findall(r"\w", w) == []:
+                    if _temp == []:
+                        continue
+                    result.append(_temp)
+                    _temp = []
+                else:
+                    _temp.append(w)
+                if i + 1 == len(text):
+                    result.append(_temp)
+            return result
     elif engine == "whitespace+newline":
-        segments = text.split()
+        segments = original_text.split()
+        if is_list_input:
+            result = []
+            _temp = []
+            for i, w in enumerate(text):
+                if (
+                    (re.findall(r"\s", w) != [] or
+                        re.findall(r"\n", w) != []) and
+                        re.findall(r"\w", w) == []
+                ):
+                    if _temp == []:
+                        continue
+                    result.append(_temp)
+                    _temp = []
+                else:
+                    _temp.append(w)
+                if i + 1 == len(text):
+                    result.append(_temp)
+            return result
     elif engine == "tltk":
         from pythainlp.tokenize.tltk import sent_tokenize as segment
-
-        segments = segment(text)
+        segments = segment(original_text)
     elif engine == "thaisum":
         from pythainlp.tokenize.thaisumcut import (
             ThaiSentenceSegmentor as segmentor,
         )
-
         segment = segmentor()
-        segments = segment.split_into_sentences(text)
+        segments = segment.split_into_sentences(original_text)
     elif engine.startswith("wtp"):
         if "-" not in engine:
             _size = "mini"
         else:
             _size = engine.split("-")[-1]
         from pythainlp.tokenize.wtsplit import tokenize as segment
-
-        segments = segment(text, size=_size, tokenize="sentence")
+        segments = segment(original_text, size=_size, tokenize="sentence")
     else:
         raise ValueError(
             f"""Tokenizer \"{engine}\" not found.
@@ -435,7 +515,12 @@ def sent_tokenize(
     if not keep_whitespace:
         segments = strip_whitespace(segments)
 
-    return segments
+    if is_list_input and engine not in ["crfcut"]:
+        word_indices = indices_words(text)
+        result = map_indices_to_words(word_indices, segments)
+        return result
+    else:
+        return [segments]
 
 
 def paragraph_tokenize(

tests/test_tokenize.py

@@ -333,6 +333,23 @@ def test_sent_tokenize(self):
         #         engine="wtp-large",
         #     ),
         # )
+        sent_4 = ["ผม", "กิน", "ข้าว", " ", "\n", "เธอ", "เล่น", "เกม"]
+        self.assertEqual(
+            sent_tokenize(sent_4, engine="crfcut"),
+            [["ผม", "กิน", "ข้าว", " ", "\n", "เธอ", "เล่น", "เกม"]],
+        )
+        self.assertEqual(
+            sent_tokenize(sent_4, engine="whitespace"),
+            [["ผม", "กิน", "ข้าว"], ["\n", "เธอ", "เล่น", "เกม"]],
+        )
+        self.assertEqual(
+            sent_tokenize(sent_4, engine="whitespace+newline"),
+            [["ผม", "กิน", "ข้าว"], ["เธอ", "เล่น", "เกม"]],
+        )
+        self.assertEqual(
+            sent_tokenize(sent_4, engine="thaisum"),
+            [["ผม", "กิน", "ข้าว", " ", "เธอ", "เล่น", "เกม"]],
+        )
         self.assertFalse(
             " "
             in sent_tokenize(