AI Topic Analysis Guidelines

Note: This is only available for the Professional Analytics license.

Our AI integration allows you to quickly analyse open answers and let AI do the work for you! The topic detection will allocate the given answer to one or mulitple topics. Below you will find the best practices that will help you to use Topic Detection effectively.

TipGood ExampleBad example
Topics should be specific enough to avoid unwanted interpretationsAppearance
(if you are aiming to identify comments related to visual aspect of the product)
Design
(could also regard usability or quality of a product)
Topics should be mutually exclusive
When topics are too closely related, AI can assign a comment to both topics in the majority of cases.
Appearance
Customer Service
Customer Service
Service
Use short specific topics instead of long onesSpeed
Punctuality
Carefulness
Reliability
Careful and reliable delivery in time
Avoid repeating the same word in multiple topicsSpeed
Punctuality
Carefulness
Reliability
Delivery Speed
Delivery Punctuality
Delivery Carefulness
Delivery Reliability
Topics should be neutral in their meaning
This way they can detect positive and negative mentions of the topic.
PriceExpensive

Technical tips

  • If all analysed open answers are in the same language, provide topics in the same language as the answers. If your answers are given in multiple languages, provide topics in English (even if you don’t have English comments).
  • Don’t use too many topics. With big topic lists the AI consistency may be lower. Try not to go over 20 topics, with 10 being a sweet spot.
  • Do not use topics like “Other” or “General. Every topic analysis already has a default “No Topic” variable that is assigned when no topic from the list can be matched to the open answer.
  • Remember that topic names are only for AI to detect them as effectively as possible. You can always do necessary adjustments directly on your dashboard:

    1. Name it differently in your dashboard so that it’s easier to understand by the report reader.
    2. Aggregate few specific topics in one general topic to track as your KPI. E.g. you can aggregate “Speed” and “Punctuality” into one aggregated “Delivery time” topic to track as your KPI on your dashboard.

AI Topics Suggestion

If you don’t want to start from scratch, you can use AI for an initial suggestionlist. It will analyse a sample of your open question answers and provide a suggested list of topics based on your configuration.
The AI-suggested list may need your manual revision in the next step so that the final topic structure meets your analytics goals and our AI topic analysis guidelines.

We also recommend copying the list to an external place for future reference, since it won’t be available again after topics are modified or AI sentiment and topics analysis feature set-up is not finished.

Updated on November 30, 2023
Was this article helpful?

Related Articles

Need Support?
Please login to your Survalyzer account and use the "Create Support Request" form.
Login to Survalyzer