Turning Curriculum Course Description into Skill Evidence: A Case Study of Two Undergraduate Programs

Abstract

Degree-program documents often help students see what they
may learn, but the short course descriptions in these
documents often state learning aims, without naming
specific skill terms that students can compare with job
requirement lists, such as tool names, method names, and
transferable work skills. When these terms are not
explicitly stated, students may need to infer what a course
covers from general wording. Maintaining up-to-date skill
summaries from course descriptions also requires staff time
and domain knowledge, because descriptions must be
interpreted consistently across courses. We address this
problem by developing a course-to-skill extractor that
converts each course description into three labeled lists:
Tools (named technologies), Hard Skills (named technical
methods), and Soft Skills (named transferable
capabilities). We apply the extractor to two undergraduate
programs (Data Science and Information Systems) using
Qwen2.5 (dense), constrained to this three-list output
format. This study makes three contributions. (1) We
provide a text-grounded workflow that extracts only the
skill terms explicitly stated in official course
descriptions and organizes them into the three categories
above. (2) We produce auditable program summaries by
storing each extracted skill label together with its
course-source, so readers can verify program-level
statements by checking the original descriptions. (3) In
this case study, we find that many skill labels appear in
only one course description and that explicit tool names
are uncommon; therefore, cross-program interpretation is
more defensible when it focuses on skill labels that recur
across multiple courses.