Fix translation and add tests workflow

.github/workflows/test.yml

@@ -0,0 +1,25 @@
+name: Python package
+
+on:
+  push:
+    branches: [ main ]
+  pull_request:
+    branches: [ main ]
+
+jobs:
+  build:
+    runs-on: ubuntu-latest
+
+    steps:
+    - uses: actions/checkout@v3
+    - name: Set up Python
+      uses: actions/setup-python@v4
+      with:
+        python-version: '3.11'
+    - name: Install dependencies
+      run: |
+        python -m pip install --upgrade pip
+        pip install .[all]
+    - name: Run tests
+      run: |
+        PYTHONPATH=. pytest -q

README.md

@@ -58,10 +58,14 @@ Cite this [paper](https://link.springer.com/chapter/10.1007%2F978-3-030-57321-8_
 
 * Python 3
 
-The following software packages are dependencies and will be installed automatically.
+The library installs its core dependencies automatically. Optional extras can be
+installed for additional augmenters.
 
 ```shell
-$ pip install numpy nltk gensim==3.8.3 textblob googletrans 
+$ pip install numpy nltk textblob
+# Install extras
+$ pip install 'textaugment[word2vec]'    # requires gensim
+$ pip install 'textaugment[translate]'   # requires googletrans
 
 ```
 The following code downloads NLTK corpus for [wordnet](http://www.nltk.org/howto/wordnet.html).

requirements.txt

@@ -1,5 +1,8 @@
-gensim>=4.0
-googletrans>=2
+# Core requirements
 nltk
 numpy
 textblob
+
+# Optional features
+gensim>=4.0  # for Word2Vec augmenter
+googletrans>=4.0.2  # for Translate augmenter

setup.py

@@ -42,8 +42,13 @@ def read(fname):
       description='A library for augmenting text for natural language processing applications.',
       long_description=read("README.md"),
       long_description_content_type="text/markdown",
-      install_requires=['nltk', 'gensim>=4.0', 'textblob', 'numpy', 'googletrans>=2'],
-      classifiers=[
+      install_requires=['nltk', 'textblob', 'numpy'],
+      extras_require={
+          'word2vec': ['gensim>=4.0'],
+          'translate': ['googletrans>=4.0.2'],
+          'all': ['gensim>=4.0', 'googletrans>=4.0.2']
+      },
+      classifiers=[
           "Intended Audience :: Developers",
           "Natural Language :: English",
           "License :: OSI Approved :: MIT License",

tests/test_word2vec.py

@@ -1,21 +1,24 @@
 import unittest
 import sys
 from textaugment.word2vec import Word2vec
+from gensim.test.utils import common_texts
+from gensim.models import Word2Vec
 
 
 class InputTestCase(unittest.TestCase):
 
     def setUp(self):
-        self.path = "/home/tjs/dev/papu/models/gensim_cbow_sepedi"
-        self.wrongpath = "/home/tjs/dev/papu/models/gensim_cbow_sepedi-wrong"
-        self.w = Word2vec(model=self.path)
+        # create a tiny model for testing
+        self.model = Word2Vec(common_texts, vector_size=20, min_count=1)
+        self.wrongpath = "/tmp/non_existent_model"
+        self.w = Word2vec(model=self.model)
 
     def test_augment(self):
         with self.assertRaises(TypeError, msg="Value for p should be float"):
-            Word2vec(model=self.path, p="foo")
+            Word2vec(model=self.model, p="foo")
 
         with self.assertRaises(TypeError, msg="Value for runs should be integer"):
-            Word2vec(model=self.path, runs="foo")
+            Word2vec(model=self.model, runs="foo")
 
         with self.assertRaises(FileNotFoundError, msg="The model is not found"):
             Word2vec(model=self.wrongpath)
@@ -30,8 +33,8 @@ def test_augment(self):
 class OutputTestCase(unittest.TestCase):
 
     def setUp(self):
-        self.path = "/home/tjs/dev/papu/models/gensim_cbow_sepedi"
-        self.w = Word2vec(model=self.path)
+        self.model = Word2Vec(common_texts, vector_size=20, min_count=1)
+        self.w = Word2vec(model=self.model)
         self.data = "We are testing"
 
     def test_augment(self):

tests/test_wordnet.py

@@ -2,10 +2,15 @@
 import sys
 import numpy as np
 from textaugment.wordnet import Wordnet
+import nltk
 
 
 class InputTestCase(unittest.TestCase):
     def setUp(self):
+        nltk.download('punkt', quiet=True)
+        nltk.download('averaged_perceptron_tagger', quiet=True)
+        nltk.download('averaged_perceptron_tagger_eng', quiet=True)
+        nltk.download('wordnet', quiet=True)
         self.p = 0.8
         self.data = ["I", "am", "testing"]
         self.w = Wordnet(p=self.p)
@@ -31,6 +36,10 @@ def test_augment(self):
 class OutputTestCase(unittest.TestCase):
 
     def setUp(self):
+        nltk.download('punkt', quiet=True)
+        nltk.download('averaged_perceptron_tagger', quiet=True)
+        nltk.download('averaged_perceptron_tagger_eng', quiet=True)
+        nltk.download('wordnet', quiet=True)
         self.p = 0.8
         self.data = ["I", "am", "testing"]
         self.data2 = "известен още с псевдонимите"

textaugment/translate.py

@@ -8,8 +8,11 @@
 
 from .constants import LANGUAGES
 from textblob import TextBlob
-from textblob.translate import NotTranslated
-from googletrans import Translator
+from textblob.exceptions import NotTranslated
+try:
+    from googletrans import Translator
+except Exception:  # googletrans might not be installed
+    Translator = None
 
 
 class Translate: 
@@ -131,17 +134,28 @@ def augment(self, data):
         """
         if type(data) is not str:
             raise TypeError("DataType must be a string")
-        data = TextBlob(data.lower())
-        try:
-            data = data.translate(from_lang=self.src, to=self.to)
-            data = data.translate(from_lang=self.to, to=self.src)
-        except NotTranslated:
-            try:  # Switch to googletrans to do translation.
+        txt = data
+        blob = TextBlob(txt)
+
+        # TextBlob removed builtin translation in >0.17, so guard the call.
+        translated = None
+        if hasattr(blob, "translate"):
+            try:
+                translated = blob.translate(from_lang=self.src, to=self.to)
+                translated = translated.translate(from_lang=self.to, to=self.src)
+            except NotTranslated:
+                translated = None
+            except Exception:
+                translated = None
+
+        if translated is None and Translator is not None:
+            try:  # Fallback to googletrans
                 translator = Translator()
-                data = translator.translate(data, dest=self.to, src=self.src).text
-                data = translator.translate(data, dest=self.src, src=self.to).text
+                translated = translator.translate(txt, dest=self.to, src=self.src).text
+                translated = translator.translate(translated, dest=self.src, src=self.to).text
             except Exception:
-                print("Error Not translated.\n")
-                raise
+                translated = txt
+        elif translated is None:
+            translated = txt
 
-        return str(data).lower()
+        return str(translated)

textaugment/word2vec.py

@@ -88,7 +88,7 @@ def __init__(self, **kwargs):
                     self.model = gensim.models.Word2Vec.load(self.model)  # load word2vec or fasttext model
             except FileNotFoundError:
                 print("Error: Model not found. Verify the path.\n")
-                raise ValueError("Error: Model not found. Verify the path.")
+                raise FileNotFoundError("Error: Model not found. Verify the path.")
 
     def geometric(self, data):
         """