Uploaded by avirupkundu411

eng pro

advertisement
AUTO-CORRECTION OF
SPELLINGS
SPANDAN DAS (G1 26)
SHOUVIK DUTTA (G2 21)
ARNAB SAHA (G1 36)
RITIKA BISWAS (G1 29)
AVIRUP KUNDU (G1 2)
PRIYANJANA DAS (G1 16)
1. What is the project all about?
The project is all about auto-correction of spellings. In this project we made
a web page which automatically check a word and correct it’s spelling.
Auto-correct is a type of software program that identifies misspelled words,
uses algorithms to identify the words most likely to have been intended,
and edits the text accordingly. Auto-correct is commonly a feature included
in word processors and messaging platforms of various types. Apple,
Google and Microsoft products all have their own versions of auto-correct
programs. auto-correct programs work very well with common and
straight-forward typos, such as "teh" for "the." However, their algorithms
are far from infallible, in particular with auto-correct programs for
smartphones, which are often more active as a means of facilitating typing
on the devices' small keyboards. There are a number of websites dedicated
to humorous auto-correct fails.
2. Why do you think your project is important?
The project is important because in this new generation most of
the people type any words or any letter or any spellings fast
because of short time and also many people who don’t know the
right spelling of a word they can easily know and type the right
spelling of the words.
Again, if you’re not a good speller, autocorrect biggest
advantage may be in saving you time. It allows you to spend less
time editing material and makes you more productive in finishing
your writing projects much more quickly. Autocorrect can take
long phrases and make them shorter. It can also create proper
fractions from common fractions. Global AutoCorrect from
LexAble software claims to make intelligent suggestions for
corrections. If you make a typo, the system analyzes its sound
patterns to determine the word you’re trying to use and corrects
immediately. If the software can’t figure out the word you’re
looking for, you can find the word and Global AutoCorrect will
use it in the future. When an autocorrect suggestion comes up
with the correct word for your context, you have just learned how
to spell the word. You might also retain the proper punctuation or
grammar you’ve learned from a sentence of your work.
3. How did you develop the project (methodology)?
Improving speed — SymSpell approach :To improve speed — let’s use an idea from SymSpell. Idea is quite
elegant. Instead of generating all possible edits each time we met
incorrect word — we can pre-calculate all delete typos (and other
typos derived from deletes). You can read more details in the original
article.
Improve memory consumption :To get the best possible accuracy we need a big dataset (at least few
gigabytes). Training n-gram model on 600 mb file leads to significant
memory consumption (25 Gb). Half of that size is used by language
model, and another half by the symspell index.
Download