Computer Science Research
Michaud specializes in computational linguistics and creating computational environments to support the language learner. Focusing on Spanish-to-English translations, she wants to see if a machine translation evaluation algorithm can be adapted to judge and respond to the accuracy of a student’s written translation.
She wants to help the algorithm figure out which words in the original Spanish sentence were translated into what English words. The current algorithm – designed for translations written by machines, not humans – fails to recognize that words which may look similar but mean something totally different in the other language may represent a very uniquely human form of mistranslation.
“The famous example is to use embarazada in Spanish to mean embarrassed, when it actually means pregnant,and that can definitely make you more embarrassed,” said Michaud.
Jim Petr ’13 will assist Michaud on a semester-long project, exploring dictionaries in Spanish and English and mining from them pairs of similarly spelled words, using a spell-check algorithm to suggest possible pairs.
These pairs can be placed into a table and that table used by the evaluation metric to recognize when a student has mistranslated a word in this way, Michaud explains.
“Our goal is to develop an algorithm to analyze a student’s translation, in order to help people work on translations and get feedback on them using online tutorial environments outside of the classroom,” she said.
Petr will add a significant project on his resume and will be the co-author of at least one article on the work.
“I am doing this project for the challenge and the chance to create a new, complete and interesting product,” he said.
Dr. Charlotte Berkes and Dr. Jimmy Franco are using a supercomputer to collaborate on research into new drugs to treat people with HIV, histoplasmosis and tuberculosis. Their results could improve the quality of life for thousands of people.
Stephanie Lapierre recently received the first Science & Engineering Dean’s Grant.