Remove duplicated weight path, package folder, object.pk

.gitignore

@@ -99,3 +99,6 @@ ENV/
 
 # mypy
 .mypy_cache/
+
+# macOS
+*.DS_Store

MANIFEST.in

@@ -5,4 +5,3 @@ include README.rst
 # Include the data files
 # recursive-include deepcut *
 include deepcut/weight/best_cnn3.h5
-include deepcut/weight/object.pk

README.md

@@ -1,5 +1,7 @@
 # Deepcut
 
+[![License](https://img.shields.io/badge/license-MIT-blue.svg?style=flat)](https://github.com/rkcosmos/deepcut/blob/master/LICENSE)
+
 A Thai word tokenization library using Deep Neural Network.
 
 ## What's new on version 0.5.1.0

deepcut/train.py

@@ -103,8 +103,8 @@ def prepare_feature(best_processed_path, option='train'):
     df = pd.concat(df)
     df = pd.concat((df_pad, df, df_pad)) # pad with empty string feature
 
-    df['char'] = df['char'].map(lambda x: CHARS_MAP.get(x, 0))
-    df['type'] = df['type'].map(lambda x: CHAR_TYPES_MAP.get(x, 0))
+    df['char'] = df['char'].map(lambda x: CHARS_MAP.get(x, 80))
+    df['type'] = df['type'].map(lambda x: CHAR_TYPES_MAP.get(x, 4))
     df_pad = create_n_gram_df(df, n_pad=n_pad)
 
     char_row = ['char' + str(i + 1) for i in range(n_pad_2)] + \

requirements.txt

@@ -1,3 +1,5 @@
-keras>=2.0.0
-tensorflow
+numpy
 pandas
+scikit-learn
+tensorflow
+keras>=2.0.0

setup.py

@@ -1,22 +1,25 @@
+#! /usr/bin/env python
 """
 Thai word Segmentation using Convolutional Neural Network
 """
 
 from setuptools import setup
 
 setup(
-  name = 'deepcut',
-  packages = ['deepcut'],
-  include_package_data=True,
-  version = '0.5.1.0',
-  install_requires=['tensorflow', 'keras>=2.0.0', 'pandas',
-                    'scipy', 'numpy', 'scikit-learn', 'h5py'],
-  license='MIT',
-  description = 'A Thai word tokenization library using Deep Neural Network',
-  author = 'Rakpong Kittinaradorn',
-  author_email = 'r.kittinaradorn@gmail.com',
-  url = 'https://github.com/rkcosmos/deepcut',
-  download_url = 'https://github.com/rkcosmos/deepcut/package/0.5.zip',
-  keywords = ['thai word segmentation deep learning neural network development'],
-  classifiers = ['Development Status :: 3 - Alpha'],
+    name='deepcut',
+    packages=['deepcut'],
+    include_package_data=True,
+    version='0.5.1.0',
+    install_requires=['tensorflow', 'keras>=2.0.0', 'pandas',
+                      'scipy', 'numpy', 'scikit-learn', 'h5py'],
+    license='MIT',
+    description='A Thai word tokenization library using Deep Neural Network',
+    author='Rakpong Kittinaradorn',
+    author_email='r.kittinaradorn@gmail.com',
+    url='https://github.com/rkcosmos/deepcut',
+    download_url='https://github.com/rkcosmos/deepcut.git',
+    keywords=['thai word segmentation deep learning neural network development'],
+    classifiers=[
+        'Development Status :: 3 - Alpha'
+    ],
 )