plane 0.2.0

A lib for text preprocessing

Plane

Plane is a tool for shaping wood using muscle power to force the cutting blade over the wood surface.
from Wikipedia

This package is used for extracting or replacing specific parts from text, like URL, Email, HTML tags, telephone numbers and so on. Also supports punctuation normalization and removement.

See the full Documents.

Install

Python 3.x only.

pip

pip install plane

Install from source

python setup.py install

Features

• no other dependencies
• build-in regex patterns: plane.pattern.Regex
• custom regex patterns
• pattern combination
• extract, replace patterns
• segment sentence
• chain function calls: plane.plane.Plane
• pipeline: plane.Pipeline

Usage

Quick start

Use regex to extract or replace:

from plane import EMAIL, extract, replace
text = '[e-mail] & [e-mail]'

emails = extract(text, EMAIL) # this return a generator object
for e in emails:
    print(e)

>>> Token(name='Email', value='[e-mail]', start=0, end=11)
>>> Token(name='Email', value='[e-mail]', start=14, end=34)

print(EMAIL)

>>> Regex(name='Email', pattern='([a-zA-Z0-9_.+-]+@[a-zA-Z0-9-]+\\.[a-zA-Z0-9-]+)', repl='<Email>')

replace(text, EMAIL) # replace(text, Regex, repl), if repl is not provided, Regex.repl will be used

>>> '<Email> & <Email>'

replace(text, EMAIL, '')

>>> ' & '

pattern

Regex is a namedtuple with 3 items:

• name
• pattern: Regular Expression
• repl: replacement tag, this will replace matched regex when using replace function

# create new pattern
from plane import build_new_regex
custom_regex = build_new_regex('my_regex', r'(\d{4})', '<my-replacement-tag>')

Also, you can build new pattern from default patterns.

Attention: this should only be used for language range.

from plane import extract, build_new_regex, CHINESE_WORDS
ASCII = build_new_regex('ascii', r'[a-zA-Z0-9]+', ' ')
WORDS = ASCII + CHINESE_WORDS
print(WORDS)

>>> Regex(name='ascii_Chinese_words', pattern='[a-zA-Z0-9]+|[\\U00004E00-\\U00009FFF\\U00003400-\\U00004DBF\\U00020000-\\U0002A6DF\\U0002A700-\\U0002B73F\\U0002B740-\\U0002B81F\\U0002B820-\\U0002CEAF\\U0002CEB0-\\U0002EBEF]+', repl=' ')

text = "自然语言处理太难了！who can help me? (╯▔🔺▔)╯"
print(' '.join([t.value for t in list(extract(text, WORDS))]))

>>> "自然语言处理太难了 who can help me"

from plane import CHINESE, ENGLISH, NUMBER
CN_EN_NUM = sum([CHINESE, ENGLISH, NUMBER])
text = "佛是虚名，道亦妄立。एवं मया श्रुतम्। 1999 is not the end of the world. "
print(' '.join([t.value for t in extract(text, CN_EN_NUM)]))

>>> "佛是虚名，道亦妄立。 1999 is not the end of the world."

Default Regex: Details

• URL: only ASCII
• EMAIL: local-part@domain
• TELEPHONE: like xxx-xxxx-xxxx
• SPACE: , \t, \n, \r, \f, \v
• HTML: HTML tags, Script part and CSS part
• ASCII_WORD: English word, numbers, <tag> and so on.
• CHINESE: all Chinese characters (only Han and punctuations)
• CJK: all Chinese, Japanese, Korean(CJK) characters and punctuations
• THAI: all Thai and punctuations
• VIETNAMESE: all Vietnames and punctuations
• ENGLISH: all English chars and punctuations
• NUMBER: 0-9

Regex name	replace
URL	'<URL>'
EMAIL	'<Email>'
TELEPHONE	'<Telephone>'
SPACE	' '
HTML	' '
ASCII_WORD	' '
CHINESE	' '
CJK	' '

segment

segment can be used to segment sentence, English and Numbers like 'PS4' will be keeped and others like Chinese '中文' will be split to single word format ['中', '文'].

from plane import segment
segment('你看起来guaiguai的。<EOS>')
>>> ['你', '看', '起', '来', 'guaiguai', '的', '。', '<EOS>']

punctuation

punc.remove will replace all unicode punctuations to ' ' or something you send to this function as paramter repl. punc.normalize will normalize some Unicode punctuations to English punctuations.

Attention: '+', '^', '$', '~' and some chars are not punctuation.

from plane import punc

text = 'Hello world!'
punc.remove(text)

>>> 'Hello world '

# replace punctuation with special string
punc.remove(text, '<P>')

>>> 'Hello world<P>'

# normalize punctuations
punc.normalize('你读过那本《边城》吗？什么编程？！人生苦短，我用 Python。')

>>> '你读过那本(边城)吗?什么编程?!人生苦短,我用 Python.'

Chain function

Plane contains extract, replace, segment and punc.remove, punc.normalize, and these methods can be called in chain. Since segment returns list, it can only be called in the end of the chain.

Plane.text saves the result of processed text and Plane.values saves the result of extracted strings.

from plane import Plane
from plane.pattern import EMAIL

p = Plane()
p.update('My email is [e-mail].').replace(EMAIL, '').text # update() will init Plane.text and Plane.values

>>> 'My email is .'

p.update('My email is [e-mail].').replace(EMAIL).segment()

>>> ['My', 'email', 'is', '<Email>', '.']

p.update('My email is [e-mail].').extract(EMAIL).values

>>> [Token(name='Email', value='[e-mail]', start=12, end=24)]

Pipeline

You can use Pipeline if you like.

segment and extract can only present in the end.

from plane import Pipeline, replace, segment
from plane.pattern import URL

pipe = Pipeline()
pipe.add(replace, URL, '')
pipe.add(segment)
pipe('http://www.guokr.com is online.')

>>> ['is', 'online', '.']