bensalilijames / normalizer

Normalize, clean and fix text

npm install node-normalizer

The simple app processes input and tries to make it consumable for a bot.

A single method clean(message) is exposed.

import normalizer from node-normalizer;
// const normalizer = require('node-normalizer').default;
const cleanedMessage = normalizer.clean('my message');

The order in which the processing happens is important.

• <xxx means sentence start then xxx
• 1. spelling corrections for common spelling errors
• 2. idiom conversions
• 3. junk word removal from sentence
• 5. special sentence effects (question, exclamation, revert question)
• 6. abbreviation expansion and canonization
• for abbreviations, do not use _ before the .
• for apostrophied left side, must follow tokenizing conventions
• for apostrophied right side, it means do not spell check the word, the apostrophe will disappear
• Format is left phrase separated by _ yields right phrase separated by +
• if right side is %value means set that bit on the sentence (%EXCLAMATIONMARK %QUESTIONMARK)
• if right side is a ~word its an interjection
• only proper names should have capital letters
• Right phrase missing means delete left phrase
• Substitutions files include:
• we use + because we dont want the resulting phrase recognized by the idiom processor and thus cause the processor to delete the phrase
• xxx> means sentence then end stop
• if you want to have the result NOT tokenized, put it in quotes