Deriving Insights From Open-Ended Learner Feedback: An Exploration of Natural Language Processing Approaches
Bibliographic record
Abstract
INTRODUCTION: Open-ended feedback from learners offers valuable insights for adapting continuing health education to their needs; however, this feedback is burdensome to analyze with qualitative methods. Natural language processing offers a potential solution, but it is unclear which methods provide useful insights. We evaluated natural language processing methods for analyzing open-ended feedback from continuing professional development training at a psychiatric hospital. METHODS: The data set consisted of survey responses from staff participants, which included two text responses on how participants intended to use the training ("intent to use"; n = 480) and other information they wished to share ("open-ended feedback"; n = 291). We analyzed "intent-to-use" responses with topic modeling, "open-ended feedback" with sentiment analysis, and both responses with large language model (LLM)-based clustering. We examined outputs of each approach to determine their value for deriving insights about the training. RESULTS: Our results indicated that because the "intent-to-use" responses were short and lacked diversity, topic modeling was not useful in differentiating content between the topics. For "open-ended feedback," sentiment scores did not accurately reflect the valence of responses. The LLM-based clustering approach generated meaningful clusters characterized by semantically similar words for both responses. DISCUSSION: LLMs may be a useful approach for deriving insights from learner feedback because they capture context, making them capable of distinguishing between responses that use similar words to convey different topics. Future directions involve exploring other methods involving LLMs, or examining how these methods fare on other data sets or types of learner feedback.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.087 | 0.183 |
| Meta-epidemiology (narrow) | 0.002 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.002 |
| Bibliometrics | 0.006 | 0.003 |
| Science and technology studies | 0.002 | 0.005 |
| Scholarly communication | 0.008 | 0.011 |
| Open science | 0.004 | 0.006 |
| Research integrity | 0.002 | 0.003 |
| Insufficient payload (model declined to judge) | 0.002 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".