neattext

NeatText:a simple NLP package for cleaning textual data and text preprocessing. Simplifying Text Cleaning For NLP & ML

Problem

• Cleaning of unstructured text data
• Reduce noise [special characters,stopwords]
• Reducing repetition of using the same code for text preprocessing

Solution

• convert the already known solution for cleaning text into a reuseable package

Docs

• Check out the full docs here

Installation

pip install neattext

Usage

• The OOP Way(Object Oriented Way)
• NeatText offers 5 main classes for working with text data
  • TextFrame : a frame-like object for cleaning text
  • TextCleaner: remove or replace specifics
  • TextExtractor: extract unwanted text data
  • TextMetrics: word stats and metrics
  • TextPipeline: combine multiple functions in a pipeline

Overall Components of NeatText

Using TextFrame

• Keeps the text as TextFrame object. This allows us to do more with our text.
• It inherits the benefits of the TextCleaner and the TextMetrics out of the box with some additional features for handling text data.
• This is the simplest way for text preprocessing with this library alternatively you can utilize the other classes too.

>>> import neattext as nt 
>> mytext = "This is the mail [email protected] ,our WEBSITE is https://example.com 😊."
>>> docx = nt.TextFrame(text=mytext)
>>> docx.text 
"This is the mail [email protected] ,our WEBSITE is https://example.com 😊."
>>>
>>> docx.describe()
Key      Value          
Length  : 73             
vowels  : 21             
consonants: 34             
stopwords: 4              
punctuations: 8              
special_char: 8              
tokens(whitespace): 10             
tokens(words): 14             
>>> 
>>> docx.length
73
>>> # Scan Percentage of Noise(Unclean data) in text
>>> d.noise_scan()
{'text_noise': 19.17808219178082, 'text_length': 73, 'noise_count': 14}
>>> 
>>> docs.head(16)
'This is the mail'
>>> docx.tail()
>>> docx.count_vowels()
>>> docx.count_stopwords()
>>> docx.count_consonants()
>>> docx.nlongest()
>>> docx.nshortest()
>>> docx.readability()

Basic NLP Task (Tokenization,Ngram,Text Generation)

>>> docx.word_tokens()
>>>
>>> docx.sent_tokens()
>>>
>>> docx.term_freq()
>>>
>>> docx.bow()

Basic Text Preprocessing

>>> docx.normalize()
'this is the mail [email protected] ,our website is https://example.com 😊.'
>>> docx.normalize(level='deep')
'this is the mail examplegmailcom our website is httpsexamplecom '

>>> docx.remove_puncts()
>>> docx.remove_stopwords()
>>> docx.remove_html_tags()
>>> docx.remove_special_characters()
>>> docx.remove_emojis()
>>> docx.fix_contractions()
>>> docx.remove_accents()
>>> docx.remove_non_ascii()

Handling Files with NeatText

• Read txt file directly into TextFrame

>>> import neattext as nt 
>>> docx_df = nt.read_txt('file.txt')

• Alternatively you can instantiate a TextFrame and read a text file into it

>>> import neattext as nt 
>>> docx_df = nt.TextFrame().read_txt('file.txt')

Chaining Methods on TextFrame

>>> t1 = "This is the mail [email protected] ,our WEBSITE is https://example.com 😊 and it will cost $100 to subscribe."
>>> docx = TextFrame(t1)
>>> result = docx.remove_emails().remove_urls().remove_emojis()
>>> print(result)
'This is the mail  ,our WEBSITE is   and it will cost $100 to subscribe.'

Clean Text

• Clean text by removing emails,numbers,stopwords,emojis,etc
• A simplified method for cleaning text by specifying as True/False what to clean from a text

>>> from neattext.functions import clean_text
>>> 
>>> mytext = "This is the mail [email protected] ,our WEBSITE is https://example.com 😊."
>>> 
>>> clean_text(mytext)
'mail [email protected] ,our website https://example.com .'

• You can remove punctuations,stopwords,urls,emojis,multiple_whitespaces,etc by setting them to True.

• You can choose to remove or not remove punctuations by setting to True/False respectively

>>> clean_text(mytext,puncts=True)
'mail example@gmailcom website https://examplecom '
>>> 
>>> clean_text(mytext,puncts=False)
'mail [email protected] ,our website https://example.com .'
>>> 
>>> clean_text(mytext,puncts=False,stopwords=False)
'this is the mail [email protected] ,our website is https://example.com .'
>>> 

• You can also remove the other non-needed items accordingly

>>> clean_text(mytext,stopwords=False)
'this is the mail [email protected] ,our website is https://example.com .'
>>>
>>> clean_text(mytext,urls=False)
'mail [email protected] ,our website https://example.com .'
>>> 
>>> clean_text(mytext,urls=True)
'mail [email protected] ,our website .'
>>> 

Removing Punctuations [A Very Common Text Preprocessing Step]

• You remove the most common punctuations such as fullstop,comma,exclamation marks and question marks by setting most_common=True which is the default
• Alternatively you can also remove all known punctuations from a text.

>>> import neattext as nt 
>>> mytext = "This is the mail [email protected] ,our WEBSITE is https://example.com 😊. Please don't forget the email when you enter !!!!!"
>>> docx = nt.TextFrame(mytext)
>>> docx.remove_puncts()
TextFrame(text="This is the mail example@gmailcom our WEBSITE is https://examplecom 😊 Please dont forget the email when you enter ")

>>> docx.remove_puncts(most_common=False)
TextFrame(text="This is the mail examplegmailcom our WEBSITE is httpsexamplecom 😊 Please dont forget the email when you enter ")

Removing Stopwords [A Very Common Text Preprocessing Step]

• You can remove stopwords from a text by specifying the language. The default language is English
• Supported Languages include English(en),Spanish(es),French(fr)|Russian(ru)|Yoruba(yo)|German(de)

>>> import neattext as nt 
>>> mytext = "This is the mail [email protected] ,our WEBSITE is https://example.com 😊. Please don't forget the email when you enter !!!!!"
>>> docx = nt.TextFrame(mytext)
>>> docx.remove_stopwords(lang='en')
TextFrame(text="mail [email protected] ,our WEBSITE https://example.com 😊. forget email enter !!!!!")

Remove Emails,Numbers,Phone Numbers,Dates,Btc Address,VisaCard Address,etc

>>> print(docx.remove_emails())
>>> 'This is the mail  ,our WEBSITE is https://example.com 😊.'
>>>
>>> print(docx.remove_stopwords())
>>> 'This mail [email protected] ,our WEBSITE https://example.com 😊.'
>>>
>>> print(docx.remove_numbers())
>>> docx.remove_phone_numbers()
>>> docx.remove_btc_address()

Remove Special Characters

>>> docx.remove_special_characters()

Remove Emojis

>>> print(docx.remove_emojis())
>>> 'This is the mail [email protected] ,our WEBSITE is https://example.com .'

Remove Custom Pattern

• You can also specify your own custom pattern, incase you cannot find what you need in the functions using the remove_custom_pattern() function

>>> import neattext.functions as nfx 
>>> ex = "Last !RT tweeter multiple &#7777"
>>> 
>>> nfx.remove_custom_pattern(e,r'&#\d+')
'Last !RT tweeter multiple  '

Replace Emails,Numbers,Phone Numbers

>>> docx.replace_emails()
>>> docx.replace_numbers()
>>> docx.replace_phone_numbers()

Chain Multiple Methods

>>> t1 = "This is the mail [email protected] ,our WEBSITE is https://example.com 😊 and it will cost $100 to subscribe."
>>> docx = TextCleaner(t1)
>>> result = docx.remove_emails().remove_urls().remove_emojis()
>>> print(result)
'This is the mail  ,our WEBSITE is   and it will cost $100 to subscribe.'

Using TextExtractor

• To Extract emails,phone numbers,numbers,urls,emojis from text

>>> from neattext import TextExtractor
>>> docx = TextExtractor()
>>> docx.text = "This is the mail [email protected] ,our WEBSITE is https://example.com 😊."
>>> docx.extract_emails()
>>> ['[email protected]']
>>>
>>> docx.extract_emojis()
>>> ['😊']

Using TextMetrics

• To Find the Words Stats such as counts of vowels,consonants,stopwords,word-stats

>>> from neattext import TextMetrics
>>> docx = TextMetrics()
>>> docx.text = "This is the mail [email protected] ,our WEBSITE is https://example.com 😊."
>>> docx.count_vowels()
>>> docx.count_consonants()
>>> docx.count_stopwords()
>>> docx.word_stats()
>>> docx.memory_usage()

Usage

• The MOP(method/function oriented way) Way

>>> from neattext.functions import clean_text,extract_emails
>>> t1 = "This is the mail [email protected] ,our WEBSITE is https://example.com ."
>>> clean_text(t1,puncts=True,stopwords=True)
>>>'this mail examplegmailcom website httpsexamplecom'
>>> extract_emails(t1)
>>> ['[email protected]']

• Alternatively you can also use this approach

>>> import neattext.functions as nfx 
>>> t1 = "This is the mail [email protected] ,our WEBSITE is https://example.com ."
>>> nfx.clean_text(t1,puncts=True,stopwords=True)
>>>'this mail examplegmailcom website httpsexamplecom'
>>> nfx.extract_emails(t1)
>>> ['[email protected]']

Explainer

• Explain an emoji or unicode for emoji
  • emoji_explainer()
  • emojify()
  • unicode_2_emoji()

>>> from neattext.explainer import emojify
>>> emojify('Smiley')
>>> '😃'

>>> from neattext.explainer import emoji_explainer
>>> emoji_explainer('😃')
>>> 'SMILING FACE WITH OPEN MOUTH'

>>> from neattext.explainer import unicode_2_emoji
>>> unicode_2_emoji('0x1f49b')
	'FLUSHED FACE'

Usage

• The Pipeline Way

>>> from neattext.pipeline import TextPipeline
>>> t1 = """This is the mail [email protected] ,our WEBSITE is https://example.com 😊. This is visa 4111 1111 1111 1111 and bitcoin 1BvBMSEYstWetqTFn5Au4m4GFg7xJaNVN2 with mastercard 5500 0000 0000 0004. Send it to PO Box 555, KNU"""

>>> p = TextPipeline(steps=[remove_emails,remove_numbers,remove_emojis])
>>> p.fit(t1)
'This is the mail  ,our WEBSITE is https://example.com . This is visa     and bitcoin BvBMSEYstWetqTFnAumGFgxJaNVN with mastercard    . Send it to PO Box , KNU'

• Check For steps and named steps

>>> p.steps
>>> p.named_steps

• Alternatively you can also use this approach

Documentation

Please read the documentation for more information on what neattext does and how to use is for your needs.You can also check out our readthedocs page here

More Features To Add

• basic nlp task
• currency normalizer
• neattext-cli

Acknowledgements

• Inspired by packages like clean-text from Johannes Fillter and textify by JCharisTech

NB

• Contributions Are Welcomed
• Notice a bug, please let us know.
• Thanks A lot

By

• Jesse E.Agbe(JCharis)
• Jesus Saves @JCharisTech