Widespread adoption of translation technology throughout the federal government requires agencies to devote time and resources to technology insertion. There are technological and logistical hurdles that must be overcome to procure the relevant technology, install it in the workplace, make it accessible to users, and train users so that they can benefit from the technology. This talk will present findings from studies on human use of machine translation (MT) output and will provide recommendations for leveraging these findings to guide tool insertion and use in operational contexts. Analyses compare comprehension performance among different types of translations, and also examine the relationship between comprehension scores and automatically-generated methods of evaluating MT output (e.g., BLEU scores). We will discuss implications of our findings for adoption of MT resources.
Erica B. Michael is a Senior Research Scientist at the University of Maryland Center for Advanced Study of Language (CASL). She received her PhD in cognitive psychology from Penn State University. Prior to joining CASL in 2005, Dr. Michael received postdoctoral training at Carnegie Mellon and was a visiting assistant professor at Bryn Mawr College. Her research interests include lexical and semantic processing in bilinguals, and her CASL work focuses on cognitive processing in tasks such as translation and summarization.
Petra Bradley is an Associate Research Scientist at the University of Maryland Center for Advanced Study of Language (CASL). She holds a PhD in cognitive psychology with a concentration in metacognition and memory. Her research interests include human performance evaluation, judgment and decision making, performance factors, and metrics for analysis. Dr. Bradley’s research is known for its impact on civilian and military language and intelligence analysts.