IBM’s machine translation (MT) ecosystem embraces about 30 languages across multiple MT providers and even more translation vendors in order to post-edit the MT. So how do we know when MT is helping? We will explain various analytical methods used to assess MT output. For example, edit distance is an example of an MT metric that is successfully utilized in multiple geographies and configurations. An analytical feedback loop is key to enable continuous monitoring of the noise (faults) across the MT ecosystem. Combining data, analytics and an end-to-end process allows operational teams to control the chaos of linguistic noise across all machine and human translation resources while bringing value to our customers.
Surfing the Chaos of Linguistic Noise