Translating texts from one language to another is quite a complex task. However, the technology explosion over the past few decades has allowed researchers to discover and implement numerous ways to achieve this human-oriented task with a computer. Among Machine Translation (MT) developers, there has been the assumption that MT systems are good enough to support people performing certain applications in the real world. More recently, informal reports from operational and field settings have described successful, but carefully limited, use of MT output in real-world tasks. Consequently, there is certainly due need for solid evaluation methods for translation based on document utility. Several strategies have been proposed to evaluate translations using algorithms that provide an assessment of translation quality through an automated metric score. Since, several considerations have forced the field to rely heavily on these metrics, we would like to know whether there is a relationship between the popular, strictly text-based quality metrics and the end-to-end (machine and user) effectiveness metrics of concern to real users.

We study the relationship of task performance results and automated metrics by exploring various aspects of automated metric correlation with subject responses from the information extraction task. We have established that, there is a positive monotonic relationship between automated metrics and correct task response rate in our data; however it is not evident how strong the relationship is in the presence of other effects in the data. This work motivates the need to extend testing methods beyond correlations and to develop other uses of these metrics for a more user-centered evaluation. Thus, we are testing for whether it will be possible to leverage the collected data from the extraction experiment using it in statistical models that we build and test for their adequacy in predicting task results quickly and less expensively.

Aside from her doctoral studies, since the summer of 1998, Tate has been employed at the U.S. Army Research Laboratory (ARL). She began as a fellow in the STARS student fellowship program conducting summer research in acoustics and meteorology in the Battlefield Environment Division. In 2000, Tate transferred to the Multilingual Research Team of the Computer and Communication Sciences Division. There she works on a range of projects related to the evaluation of language technologies.

Tate is originally from Zachary, Louisiana, a small town just outside of Baton Rouge. In her spare time, she enjoys exploring her creativity through crafts, home projects, and event design and planning. Tate is actively involved in her sorority, Alpha Kappa Alpha Sorority, Inc. and serves as the team leader for her church, The People's Community Baptist Church, youth usher board. She has a passion for working with youth, mainly through academic programs encouraging them to excel in and purse opportunities in math, science, and technology. Tate is the first in her immediate family to go to college.