Inspired by the success of the NIST Speaker and Language i-Vector Challenge, the NIST OpenMT evaluation series is bringing to you the Machine Translation Challenge. The goal of the Challenge is to lower the entry bar to the evaluation to attract wider participation. The Challenge follows a similar protocol as the traditional OpenMT evaluation, but with a flexible schedule to allow maximum participation. The Challenge will be up for six months instead of just one week. A leaderboard will track the performance. Score feedback will be provided for 40% of the test set, and participants can submit up to 5 submissions per day.
The Challenge will use data from the OpenMT15 evaluation, which focuses on informal data genres (SMS/Chat and Conversational Telephone Speech) in Arabic-to-English and Chinese-to-English.
mt_poc [at] nist.gov
[ Main ]