Open comments are often the gold mine of a survey, but reading 800 verbatims burns out the HR team. Illizeo’s NLP analysis auto-classifies them by theme, sentiment and urgency so you focus on what matters.
Step-by-step #
-
Enable NLP analysis #
Tick NLP analysis when creating the survey. Supports EN, FR, IT, DE, ES.
-
Visualise themes #
Once closed, the survey shows a tag cloud: management, workload, team, environment, comp, purpose, growth.
-
Filter by sentiment #
Positive (green), neutral (grey), negative (red) verbatims. Combinable with themes.
-
Spot weak signals #
Under-represented but recurring themes deserve attention. E.g. 3 mentions of “burnout” are already alarming.
-
Quote in the debrief #
You can quote an anonymised verbatim to illustrate a point. Always check the quote doesn’t re-identify the author.
Auto-detected themes #
| Theme | Keywords |
|---|---|
| Management | manager, lead, leadership, listening |
| Workload | overload, pressure, deadlines, burnout |
| Team | colleagues, vibe, collaboration |
| Environment | office, remote, equipment, noise |
| Compensation | salary, bonus, raise, equity |
| Growth | career, promotion, mobility, training |
FAQ #
Is the analysis reliable?
> 90% accuracy on sentiment, > 85% on theme classification. Always combined with human review for sensitive decisions.
Does my data go to an external LLM?
No, NLP runs on Illizeo infrastructure in Switzerland. No leak to third-party LLMs.
Can I customise themes?
Yes, add your own themes (e.g. “internal tools” with keywords) or fine-tune the model on your verbatims.
Is anonymity preserved?
Yes. Analysis runs on de-identified verbatims. Names and identifiers are stripped upstream.
