Project is a webpage for English to Hindi Transliteration along with Predictive typing. Prediction is based upon the frequency of words, More the frequency, higher up the list it will appear.
Input
Initially Input is the Hindi Words, typed in English Characters.
When Suggestions are shown, user gives input in various form.
Mouse over suggestion box: The Word is Highlighted.
Mouse Click Over Suggestion: Word is Completed.
Arrow Keys: Browse Suggestions.
Space or Tab or Enter Key: Word is Completed.
Escape Key: Suggestion box is closed.
Pressing some key again: Suggestion is shows if available.
Click on Display Help: Mapping is shown.
Click on Save Words: Words are saved(if not in databse) and result is shown.
Output
Text converted to Hindi is shown in output area.
If suggestion is available while typing, suggestions are shown.
Appropriate action is taken depending on user input in suggestion box.
Problem Statement
Screenshots showing Transliteration, Clicking of button "Save All New Words" and "Result on button itself"
Motivation
Motivation/Context
I was inspired by CFILT, and had(& still have!) interest field of NLP. This project gave me a little more insight into Processing the words and complexities involved in it.
Suplemented with a database of Hindi Words, it can be very useful for editing Hindi text. Although transliteration tools are available for editing, they don't show predictions.
Motivation for implementing it as a WebPage rather than some plugin was - so that it can be available to more no. of people.
Features [(Implemented) & !(Implemented)]
Guaranteed
English to Hindi Transliteration.
Prediction.
Extra Features
Temporary Saving of words.
Not Implemented
Frequency based sugegstion.
Why? : Prof. Siva told me to try to implement it without server interaction.
Timeline
First Stage
Analyzing the topics : August 7 - August 15
Finalize the topic - August 15
Study the present resources and create first stage report - August 15 -August 23
Submission of the first stage report - August 23
Second Stage Report
Study of already existing resources on itrans : August 23 - August 31
Interface Design : September 1 - Sepetmber 5
Implement Transliteration : September 5 - September 11
Implement Prediction : September 18 - September 25
Testing/Last Min. Debugging And Wishlist : September 26 - September 28
Submission of second stage report : September 29
Solution Design
Modules: Original Work
Convert: Implements Transliteration
Predict: Implements Showing of suggestions and auto-completeing them
Animation: Implements various annimations and effects
CSS Styles: Stylesheet for enhancing graphics.
Modules: Reused
JQuery: A javascript library, used for animations
Implementation Details
Technologies Used
HTML
JavaScript
CSS
Tools Used
GEdit
Notepad++
CSSed
Firebug: Firefox Plugin
Color Picker: Firefox Plugin
Code Re-used
Functions Set/Get Caret Position in input box from internet.
Escaping Characters for RegExp in Javascript from internet.
Lines of Code
Approx ~ 1 kLoc
Challenges
Challenges Experienced
Creating the most intutive mapping of english characters to hindi unicodes.
Handling special symbols and numbers, and to escape RegExp characters.
Solutions
Mapping is almost same as they sound while we talk.
Appropriate RegExp were made and numbers/symbols were replaced before everything else.
Conclusions And Future Work
Conclusions
Learned to handle Unicodes.
Few functionalities are still lacking but it performs basic task.
Future Work
With appropriate mapping, it can be used for other languages too.
Exisiting dictionaries can be used to enhance suggestions.
