“If not for the efforts of the community, who made data and dictionaries available with minimal license restrictions,” Dendi says, “this Haitian Creole machine translator would not be available.” Government agencies released parallel documents and glossaries, and Web sites such as CrisisCommons and were happy to share glossaries and translation resources.” “For instance,” Quirk says, “Carnegie Mellon University had a repository for parallel Haitian Creole and English spoken and text data. That was when they discovered other groups who had made language resources available. The NLP group at work on the Haitian Creole translation system.īut team members quickly replaced skepticism with dogged determination and reached out for help. Compared with more widely spoken languages, the amount of parallel data for Creole is fairly limited. Approximately 8 million people in Haiti speak Creole. Haitian Creole, or Kreyòl, is one of two official languages spoken in Haiti French is the other. The NLP team knew that its biggest challenge would be identifying parallel data between English and Haitian Creole for training the engine. Statistical machine translation has the incredible ability to turn parallel translated data into translation systems in a matter of hours or days-once you have enough training data.” “When Vikram first told us the aid community had asked for a Haitian Creole machine-translation system, I was intrigued but skeptical. Driven by the urgency of the situation, Dendi’s product team and NLP researchers put aside other priorities and brainstormed ways to get an experimental but functional Haitian Creole machine-translation system online quickly.Ĭhris Quirk, a researcher with the Machine Translation team, recalls those initial meetings. Normally, adding a new language to the machine-translation engine can take weeks, if not months. Microsoft Translator’s Haitian Creole widget.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |