improve text cleaning

hexlib/regex.py

@@ -3,4 +3,4 @@ import re
 LINK_RE = re.compile(r"(https?://[\w\-_.]+\.[a-z]{2,4}([^\s<'\"]*|$))")
 HTML_HREF_RE = re.compile(r"href=\"([^\"]+)\"")
 WHITESPACE_RE = re.compile(r"\s+")
-PUNCTUATION_RE = re.compile(r"[.,;:\"']+")
+PUNCTUATION_RE = re.compile(r"[.,;:\"!?/()|*=]+")

hexlib/text.py

@@ -1,43 +1,65 @@
+from functools import partial
+from multiprocessing.pool import Pool
+
 import nltk.corpus
 from lxml import etree
 from nltk.corpus import stopwords
 from nltk.stem import WordNetLemmatizer
 
-from .regex import WHITESPACE_RE, PUNCTUATION_RE
+from .regex import WHITESPACE_RE, PUNCTUATION_RE, LINK_RE
 
 get_text = etree.XPath("//text()")
 
 stop_words_en = set(stopwords.words("english"))
 
+extra_stop_words_en = [
+    "u", "&", "-", "--"
+]
+
+stop_words_en.update(extra_stop_words_en)
+
 nltk.download("stopwords", quiet=True)
 nltk.download("wordnet", quiet=True)
 
 lemmatizer = WordNetLemmatizer()
 
 
-def clean(text, compress_whitespace=False, lowercase=False, clean_html=False, strip=False, remove_punctuation=False,
-          remove_stopwords_en=False, lemmatize=False):
-    if compress_whitespace and remove_stopwords_en:
-        raise ValueError("Redundant flags: remove_stopwords implies compress_whitespace")
+def clean_multicore(texts, processes, **kwargs):
+    pool = Pool(processes=processes)
+    return pool.map(
+        func=partial(clean, **kwargs),
+        iterable=texts,
+    )
+
+
+def clean(text, lowercase=False, clean_html=False, strip=False, remove_punctuation=False,
+          remove_stopwords_en=False, lemmatize=False, fix_single_quotes=False, strip_quotes=False,
+          remove_urls=False):
+    if fix_single_quotes:
+        text = text.replace("`", "'")
+
+    if remove_urls:
+        text = LINK_RE.sub(" ", text)
 
     if clean_html:
         try:
-            root = etree.fromstring(text)
+            root = etree.fromstring(text.replace("&", ""))
             text = "".join(get_text(root))
         except:
             pass
 
+    if remove_punctuation:
+        text = PUNCTUATION_RE.sub(" ", text)
+
     if lowercase:
         text = text.lower()
 
-    if compress_whitespace:
+    if not remove_stopwords_en or not lemmatize or not strip_quotes:
         text = WHITESPACE_RE.sub(" ", text)
 
-    if strip:
-        text = text.strip()
-
-    if remove_punctuation:
-        text = PUNCTUATION_RE.sub("", text)
+    if strip_quotes:
+        words = WHITESPACE_RE.split(text)
+        text = " ".join(w.strip("\"'") for w in words)
 
     if remove_stopwords_en or lemmatize:
         words = WHITESPACE_RE.split(text)
@@ -49,4 +71,7 @@ def clean(text, compress_whitespace=False, lowercase=False, clean_html=False, st
         elif lemmatize and not remove_stopwords_en:
             text = " ".join(lemmatizer.lemmatize(w) for w in words)
 
+    if strip:
+        text = text.strip()
+
     return text

test/test_text.py

@@ -42,7 +42,6 @@ class TestText(TestCase):
             text,
             clean_html=True,
             lowercase=True,
-            compress_whitespace=True
         )
         expected = " hello, world "
 
@@ -54,7 +53,6 @@ class TestText(TestCase):
             text,
             clean_html=True,
             lowercase=True,
-            compress_whitespace=True,
             strip=True
         )
         expected = "hello, world"
@@ -67,7 +65,6 @@ class TestText(TestCase):
             text,
             clean_html=True,
             lowercase=True,
-            compress_whitespace=True,
             strip=True,
             remove_punctuation=True
         )
@@ -103,3 +100,66 @@ class TestText(TestCase):
         expected = "hello world"
 
         self.assertEqual(cleaned, expected)
+
+    def test_html_8(self):
+        text = "<div>\n Hello, \t\n<strong>a the worlds!    </strong>\n\t</div>"
+        cleaned = clean(
+            text,
+            clean_html=True,
+            lowercase=True,
+            remove_punctuation=True,
+            strip=True,
+            remove_stopwords_en=True,
+            lemmatize=True
+        )
+        expected = "hello world"
+
+        self.assertEqual(cleaned, expected)
+
+    def test_html_9(self):
+        text = "<div>\n Hello, \t\n<strong>world! it's it`s   </strong>\n\t</div>"
+        cleaned = clean(
+            text,
+            clean_html=True,
+            lowercase=True,
+            remove_punctuation=True,
+            strip=True,
+            lemmatize=True,
+            fix_single_quotes=True
+        )
+        expected = "hello world it's it's"
+
+        self.assertEqual(cleaned, expected)
+
+    def test_html_10(self):
+        text = "<div>\n Hello, \t\n<strong>world! it's it`s https://google.ca/test/abc.pdf  </strong>\n\t</div>"
+        cleaned = clean(
+            text,
+            clean_html=True,
+            lowercase=True,
+            remove_punctuation=True,
+            strip=True,
+            lemmatize=True,
+            fix_single_quotes=True,
+            remove_urls=True
+        )
+        expected = "hello world it's it's"
+
+        self.assertEqual(cleaned, expected)
+
+    def test_html_11(self):
+        text = "<div>\n Hello, \t\n<strong>world! it's it`s u & | </strong>\n\t</div>"
+        cleaned = clean(
+            text,
+            clean_html=True,
+            lowercase=True,
+            remove_punctuation=True,
+            strip=True,
+            lemmatize=True,
+            fix_single_quotes=True,
+            remove_stopwords_en=True,
+            remove_urls=True
+        )
+        expected = "hello world"
+
+        self.assertEqual(cleaned, expected)