Skip to content

Persian (Farsi) text pre processing (normalize, number, punctuation, white space, stop word & ...)

License

Notifications You must be signed in to change notification settings

karget-platform/persian-preprocess

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

1 Commit
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

persian-preprocess

Persian (Farsi) text pre processing (normalize, number, punctuation, white space, stop word & ...)

Table of contents

npm install --save persian-preprocess
const persianPreProcess = require('persian-preprocess');
Parameter Type Required Descriptiopn
text String Yes Text to process
debug Boolean Yes Debug system status
const text = 'text to process';
const debug = true;
const processedText = persianPreProcess(text, debug)
    .normalize()
    .number()
    .lowercase()
    .punctuation()
    .remove()
    .stopword()
    .emoticon()
    .whitespace();
  • Above code is just a sample of all pre process methods and because some of the methods require parameters this code won't work correctly. For complete functional sample please check Full Sample
Normalization Methods Description
normalize Normalization process
number Change numbers locale
lowercase Lowercase all characters
punctuation Remove punctuation
remove Remove selected characters
stopword Remove stop words
emoticon Remove emoticons
whitespace Remove duplicate whitespaces
Result Methods Description
toString Get processed text
toArray Get list of all words
toUnique Get list of unique words
getDebug Get pre process debug data

Normalization process

Parameter Type Required Descriptiopn
config Object No Normalization config (table below)
Configuration Type Description Sample Characters
persian boolean Normalize persian characters ﭐ ݓ ك ﻱ
english boolean Normalize english characters ᗩ ℳ Ѡ ⓡ ⒵
arabic boolean Normalize arabic characters ﷲ ﷺ
number boolean Normalize number characters ٥ ⑩
math boolean Normalize math characters ¼ ⅞
html boolean Normalize html characters   <
punctuation boolean Normalize punctuation characters ʕ ʔ ℅ ٪
special boolean Normalize special characters ᴁ lj st
  • Default value for all configurations sets is true and the normalization process will use for all of them by default
  • Setting configurations value to false will ignore normalization process for the set
// No configuration (using default value)
const processedText = persianPreProcess(text, debug).normalize();

// Default configuration (Same result with the code above)
const processedText = persianPreProcess(text, debug).normalize({
    persian: true,
    english: true,
    arabic: true,
    number: true,
    math: true,
    html: true,
    punctuation: true,
    special: true
});

// Ignore HTML characters normalization
const processedText = persianPreProcess(text, debug).normalize({
    html: false
});

Change numbers locale

Parameter Type Required Descriptiopn
language Enum: 'persian', 'english' Yes Numeric characters locale
// 0, 1 ... 9
const processedText = persianPreProcess(text, debug).number('english');

// ۰, ۱, ... ۹
const processedText = persianPreProcess(text, debug).number('persian');

Lowercase all characters

// 'Hello' => 'hello'
const processedText = persianPreProcess(text, debug).lowercase();

Remove punctuation

Parameter Type Required Descriptiopn
config Object No Punctuation removal config (table below)
Configuration Type Description Sample Characters
basic boolean or null Basic punctuations ' " \ / , ( | )
mark boolean or null Special punctuations \r \n \t \0
diacritic boolean or null Arabic diacritics ٌ ٍ ً ّ
unicode boolean or null Unicode punctuations ZERO WIDTH NON-JOINER
  • Default value for all configurations sets is true and the punctuations will remove using space character
  • Setting value to null will remove the punctuations and wont replace them with any character
  • Setting value to false will ignore punctuations removal for the set
// No configuration (using default value)
const processedText = persianPreProcess(text, debug).punctuation();

// Default configuration (Same result with the code above)
const processedText = persianPreProcess(text, debug).punctuation({
    basic: true,
    mark: true,
    diacritic: true,
    unicode: true
});

// Ignore UNICODE punctuation removal
const processedText = persianPreProcess(text, debug).punctuation({
    unicode: false
});

/**
 * Using NULL as configuration value for basic punctuations
 *
 * Using true (default value): 'in-line' > 'in line'
 * Using null                : 'in-line' > 'inline'
 */
const processedText = persianPreProcess(text, debug).punctuation({
    basic: null
});

Remove selected characters

Parameter Type Required Descriptiopn
config Object No Character removal config (table below)
Configuration Type Description Sample Characters
number boolean or null Numeric characters 0 9 ۰ ۹
persian boolean or null Persian characters آ ا ی
english boolean or null English characters A Z a z
length number Words with specific length
  • for number, persian and english configurations

    • Default value is false and the character removal process will be ignored by default
    • Setting value to true will remove all the chacters in set and replace them with space character
    • Setting value to null will remove the characters and wont replace them with any character
  • Setting length configuration will remove all words with the length equal or less than given value

/**
 * No configuration (using default value)
 * Using method with no configuration wont make any changes to text
 */
const processedText = persianPreProcess(text, debug).remove();

// Removing all characters
const processedText = persianPreProcess(text, debug).remove({
    number: true,
    persian: true,
    english: true
});

/**
 * Using NULL as configuration value for english characters
 *
 * Using true : 'in-line' > 'in line'
 * Using null : 'in-line' > 'inline'
 */
const processedText = persianPreProcess(text, debug).remove({
    english: null
});

/**
 * Using length configuration
 * 'this is a text' > 'this   text'
 */
const processedText = persianPreProcess(text, debug).remove({
    length: 2
});

Remove stop words

Parameter Type Required Descriptiopn
config Object No Stopword removal config (table below)
Configuration Type Description Sample Words
persian boolean Persian stopwords در با به
english boolean English stopwords in at on
custom string[] List of custom Words
  • for persian and english configurations

    • Default value is false and the stopwords removal process will be ignored by default
    • Setting value to true will remove all the stopwords in set
  • Setting custom configuration will remove all words in given list

/**
 * No configuration (using default value)
 * Using method with no configuration wont make any changes to text
 */
const processedText = persianPreProcess(text, debug).stopword();

// Removing all stopwords
const processedText = persianPreProcess(text, debug).stopword({
    persian: true,
    english: true
});

/**
 * Using custom list configuration
 * 'this is a text' > ' is  text'
 */
const processedText = persianPreProcess(text, debug).stopword({
    custom: ['this', 'a']
});

Remove emoticons

Parameter Required Descriptiopn
replace No Value of this parameter can only be NULL
  • Be default (calling method with no parameter) all emoticons will remove using space character
  • Setting replace value to null will remove the emoticons and wont replace them with any character
// No configuration (using default value)
const processedText = persianPreProcess(text, debug).emoticon();

/**
 * Using NULL as replace parameter value
 *
 * No parameter : 'I💓U' > 'I U'
 * Using null   : 'I💓U' > 'IU'
 */
const processedText = persianPreProcess(text, debug).emoticon(null);

Remove duplicate whitespaces

// 'this is     a  text    ' => 'this is a text '
const processedText = persianPreProcess(text, debug).whitespace();

Get processed text

/**
 * text   : '❶ text and ❶ number'
 * result : '1 text and 1 number'
 */
const stringValue = processedText.toString();

Get list of all words

/**
 * text   : '❶ text and ❶ number'
 * result : ['1', 'text', 'and', '1', 'number']
 */
const arrayList = processedText.toArray();

Get list of unique words

/**
 * text   : '❶ text and ❶ number'
 * result : ['1', 'text', 'and', 'number']
 */
const uniqueList = processedText.toUnique();

Get pre process debug data

// See Full Sample
const debugInfo = processedText.getDebug();
const text = `
    استفاده از حرف ك عربی و کاراکتر خاص ﷼ و عدد عربی ٦
    انگلیسی: using ß character and ⅜ and <  ⒄℅
    حط دوم انگلیسی: and special character: NJ
    شکلک:  😃 👦🏿 🚩 👱🏽 🍉 🏒 🚍 🥬
    انتهای متن
`;

const persianPreProcess = require('persian-preprocess');
const processedText = persianPreProcess(text, true)
    // Normalize
    .normalize()
    // Change number locale to persian
    .number('persian')
    // Lowercase all characters
    .lowercase()
    // Remove all punctuation except marks (i.e.: \n)
    .punctuation({
        mark: false
    })
    // Remove all numeric characters (not using space space character)
    .remove({
        number: null,
    })
    // Remove persian, english and two custom stop words
    .stopword({
        custom: ['حرف', 'خط']
    })
    // Remove emoticons
    .emoticon()
    // Remove duplicate whitespaces
    .whitespace();

const result = processedText.toString();
/**
استفاده عربی کاراکتر خاص ریال عدد عربی
 انگلیسی using b character
 انگلیسی special character nj
 شکلک
 انتهای متن
*/

console.log(processedText.getDebug());
/**
 * Setting debug parameter as true for persianPreProcess
 * will activate debug system and debug data will be like:
 */
{
  TOTAL: { duration: 0.054, change: -96, length: 212 },
  normilize: {
    duration: 0.018,
    change: 4,
    length: 216,
    match: [
      'ك',    'ß', '﷼',
      '٦',    '⒄', '⅜',
      '<', '℅', 'NJ'
    ]
  },
  number: { duration: 0.001, change: 0, length: 216, match: [] },
  lowercase: { duration: 0, change: 0, length: 216 },
  punctuation: {
    duration: 0,
    change: 0,
    length: 216,
    match: [ ':', '/', '<', '%' ]
  },
  remove: {
    duration: 0,
    change: -5,
    length: 211,
    match: [ '۶', '۳', '۸', '۱', '۷' ]
  },
  stopword: {
    duration: 0.028,
    change: -6,
    length: 205,
    match: [
      'and',  ' دوم ',
      ' از ', ' ک ',
      ' و ',  ' حرف ',
      ' حط '
    ]
  },
  emoticon: {
    duration: 0.002,
    change: -10,
    length: 195,
    match: [
      '😃', '👦', '🏿',
      '🚩', '👱', '🏽',
      '🍉', '🏒', '🚍',
      '🥬'
    ]
  },
  whitespace: { duration: 0.001, change: -79, length: 116 }
}
Name Description
duration Process time in millisecond
change Number of characters added or removed from Text value
length Text value length after process
match List of matched characters/words in process
git clone https://github.com/webilix/persian-preprocess.git
npm install
npm test

About

Persian (Farsi) text pre processing (normalize, number, punctuation, white space, stop word & ...)

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • JavaScript 100.0%