Bilinual Books Purpose
Bilinual helps users improve their language skills by providing books that are annotated with translation hints.
The Bilinual objective is not to translate books, but annotates them with in-context translation and language hints. Readers should not rely on the annotation but use these hints and clues for actively guessing the right translation. Bilinual takes the burden from the user to look up the word and provide the needed information through interaction/distraction free interfaces.
What is the challenge?
Words can have different meanings. For example, if you give someone directions, you would say “Turn left at the first intersection, then turn right.” However, you might use the same word in a different context and meaning: “Freedom of speech is a basic human right.” Or, for some words, even another completely different meaning: “The algorithm needs to decide which translation is correct to display the right words.”. The tool aims to eliminate the extra step of looking up words in a dictionary and finding the best translation so that the readers can instead focus on reading and enjoying the book itself.
How does Bilinual service work?
Proper annotations are selected by a machine learning algorithm that is powered by Google Word2Vec and SpaCy/Gensim/NLTK libraries.
All books are in the public domain in the U.S and are available as a part of "Project Gutenberg"
This project was not possible without these open projects and libraries:
Fasttext, SpaCy, Gensim, Numpy, Flask, NLTK WordNet, Celery, SQLite, jQuery
Feel free to contact me at bilinualcom/at/gmail/dot/com or using the form in the right top corner if you have any questions or requests.
The picture below depicts how Bilinual is working.
What are the other challenges?
We are working on speeding up the tool - a proper language model is huge and therefore requires a lot of CPU power and memory that we cannot afford if we want to keep the service free.