Higher Education Institutions are encouraged to publish their course catalogues online. Learning Outcomes (LOs) at both programme and module levels are a central facet of programme provision. Accrediting institutions use LOs to certify whether courses meet discipline specific learning requirements/standards. Therefore, the construction of LOs remains central to the teaching endeavours of academics.
This paper reports on the use of text data mining processes to extract all published LOs from three institutional catalogues and using Part of Speech (POS) tagging to compare and contrast learning outcomes across institutions. The context behind the project was the merger of the three institutions from which the LOs were drawn namely, Institute of Technology Blanchardstown (ITB), Institute of Technology Tallagh (ITT) and Dublin Institute of Technology (DIT) were in the process of merging.
Course catalogues from the three institutions were scrapped programmatically and a single HTML document was created for each course, comprising all learning outcomes associated each course. The HTML structure of the webpage was inspected to identity key common markers associated with the learning outcomes specifically. Python programming language was used in conjunction with the natural language toolkit to extract and process the LOs text resulting in a dataset of LOs and associated metadata about the course and institution.
Once this dataset was created, a POS tagger that annotates the sentences with indicative parts of speech (noun, verb, and preposition) was employed. Utilising Bloom’s Taxonomy as a framework for analysis (as the HE guidance for LOs is based on this framework) the LOs were examined. In addition, indicators of vague LOs and open-ended learning outcomes were also identified.
Comparing all the LOs extracted, the Top 20 most repeated LOs across the 3 institutions were identified and they suggest prioritisation of well-defined key skills acquisition (personal learning plans, time management and academic integrity) as part of multiple courses. LOs, which were open ended, potentially facilitate students’ creativity compared with more vague LOs, which could potentially deter learners. Word clouds were produced to identify the most frequent action verbs in use across the corpora.
The project has illustrated that critical analysis of institutional LOs could give a framework with which to analyse the quality of course design and provide insight into the care/precision taken by the academic staff when drafting learning outcomes. Course structures and course quality can be tracked by exploring LO datasets and can aid in averting vague or excessively similar LOs throughout programme provision. The dataset could also form the base from which to support academics when drafting LOs.
