Skip to content
English
Contact us
  • There are no suggestions because the search field is empty.

How the new LLM Topic Assignment works

Understand what the AI considers to assign topics and how to best fine-tune it it to your needs. Important is to leverage the topic descriptions instead of reviewing many examples.

Here are some tips about the new AI system for topic assignment that can help you better improve the performance.
Besides the text + reviewed texts, there are now a few elements that affect the quality of the AI output. For issues that you notice are prevalent throughout the project / question, I would suggest using these new features to influence the AI results (instead of solely labelling more rows).

📁 Project & Question Context

The AI now uses the following fields during training:
  • Project Title
  • Project Description
  • Question Name
  • Question Description
image.png
image.png

✅ Best Practices

  • Use descriptive titles instead of generic ones if available
    Example: Use “Feedback on Delivery Process” instead of “Q3”.
  • Add contextual info to descriptions:
    • 🏢 Lesser-known companies or terms
      E.g., “XXX is a subsidiary of YYY that focuses on manufacturing.”
    • 🔧 Technical or company-specific terms
      Define acronyms or jargon that might not be widely known.
      Generally, the AI will be able to understand technical terms even if they are only known by specialists of the field. However, if the terms are uncommon acronyms or are company-specific, it may be useful to define them
    • 😠 Sentiment guidance
      E.g., if responses are from complaints, say so — this helps the AI interpret tone or sarcasm better.

🗂️ Topic Descriptions

Topic descriptions are now a major input for the AI.
  • These are auto-generated based on context (topic name + project/question info)
  • But… they may be wrong or unclear if topic names are vague

✅ When to Update Descriptions

  • A topic is frequently misassigned or missed
  • The topic name is unclear or ambiguous
For instance, you could write a description such as:
“This topic solely covers the quality of the food. It should NOT include comments about cost or value.”

⚠️ Notes on Behavior

  • Descriptions apply on the next training round (not instantly)
  • Changing a topic label does not auto-update the description: Usually this is not an issue if it is still referring to the same topic, but if you change it to something very different, you will have an outdated and irrelevant description.
  • To generate a new AI-written description, delete the current one (leave it blank): However, note that this will also only be re-generated on the next AI update
image.png

✅ Manual Reviewing Tips

It is still possible and useful to manually review rows! However, we expect that the new AI should have a solid enough baseline to reduce the amount of manual reviewing required.
 

🔍 Where to Focus Your Time

  • ⚠️ If an issue is widespread, update the descriptions instead
  • Skip easy/obvious rows the AI got right
  • Focus on borderline or complex examples
Label fewer rows, but make them count.
(The AI now benefits more from quality over quantity.)

🚧 Work in Progress

There are a few features that are still undergoing development, so please exercise caution when using any of the information from these parts.
  • 📔 No ‘learns from’ project functionality — Currently when creating a topic collection, you can inherit from another project. However, the new AI currently does not utilize manually reviewed rows from the original project to improve the performance of the new project.
  • 🕑 Incorrect time estimates for topic assignment — The estimates shown are not updated for the new model. Hence, there may be no estimate written at some points / the estimate may be inaccurate

🧪 Feedback & Known Issues

Possible Issues:

  • 🧠 “Too many topics assigned” — Overlapping topics confuse the model. First check that they should indeed be distinct. If so, this can sometimes cause confusion for the AI if the topics appear to have similar meanings.
  • 🕳️ “Batches of blank rows” — AI failed to assign anything for many consecutive rows

What to Do:

⚠️ Report them to support@caplena.com or in the chat.