yididiyan/amharic_spell_corrector

Amharic Spelling Corrector based on SymSpell - Spelling corrector which is 1 million times faster through Symmetric Delete spelling correction algorithm

Amharic Spelling Corrector

Spelling corrector can be handy where you need to post process outputs of an OCR(Optical Character Recognition), STT(Speech to text).

Dependencies

• PySymSpell - pure Python port of SymSpell.
• BeautifulSoap

Installation

pip install -r requirements.txt

Usage

Spelling correctortion

from spell_corrector import lookup
suggestions = lookup('አሽናቆት')
suggestions[0].term
>> 'አድናቆት'

Edit Distance Calculation

The function calculates the number of edit [ insertion, update, delete ] needed to correct a word to another

from spell_corrector import calculate_distance
calculate_distance('አክብሮትና', 'አክብሮት')
>> 1

Crawling More Words

The crawler.py script crawls around 1 million frequently used Amharic words on the Web.

python crawler.py

amharic_spell_corrector's Issues

Limited Words List

I find the words limited. I want contribute more.

Some un-spaced words in the database

First of all thank you for your crawler and the database you have made available to the public. Just one quick thing I noticed while I was working with the data. Some words aren't separated by space and still hold meanings from the site which they were scrapped from like
"ግዜሳይረፍድይህንንየቁርጥቀንኢትዮጵያዊለመታደግበያለንበትበጋራእንረባረብ
የየመንኤምባሲዎችንበያሉበትበተቃውሞማጨናነቅናእንቅልፍመንሳትእንዲሁምአለምአቀፍጫናማድረግያስፈልጋል
የመንአቶአንዳርጋቸውንለወያኔአሳልፋእንዳትሰጥናያለምንምቅድመሁኔታእንድትለቅሁሉምነጻነትናፋቂኢትዮጵያዊሁሉበጋራጥረትሊያደርግይገባል
ማሰርናማንገገላታትአለምአቀፍህግንየሚጥስህገወጥድርጊትነው
የአንባገነኑየህወሃትስርወመንግስትተቃዋሚዎችንማንምአገርበምንምአይነትሁኔታማፈን
የመንአቶአንዳርጋቸውጽጌንማገቷሌላአደገኛናበጣምአሳሳቢክስተትነው
"
I thought you might be interested to know.