Psychological applications of human language technology combined with multidisciplinary approaches to similarity calculations and data visualization offer avenues to broaden the use of students' own words in program assessment. We compared multiple analysis approaches on both simple token counts (word roots and character trigrams) and top-down language indicators from 85 student essays about service-learning events. Bioinformatic distance calculations on word root counts provided useable assessment information on attitude change, showing patterns of word use that match the holistic goals of the assignment. Although these patterns were not found in a subsequent batch of 81 essays, the tools we are providing may facilitate other efforts to detect attitude change in student writing about service-learning events.