Explorer: Ask Relevance

When dealing with a big amount of data, clustering helps to break the data into meaningful categories. A useful next step is to get a summary over each category. Relevance AI provides you with the ability to automatically generate a summary over each category on the Explorer dashboard. So make sure a category view tab/page is set up and selected.

📘

Ask Relevance on the whole data

Under category view, Ask Relevance explores items within a specified category. If you wish to ask general question regardless of categories, you can use the Ask Relevance view

What is Ask Relevance on the Explorer Category View

"Ask Relevance" on the Explorer dashboard works best for AI-Clustering results. This feature looks into the data in a selected category and provides you with an answer to your question. Questions can be written in natural language such as

• summarize this
• What are the top three main complaints by clients

How to use Ask Relevance on the Explorer Category view

On a category view tab, scroll down to the desired category and click on "Ask Relevance" on the top right of the category as shown in the image below.

Complete the setup:

• Select the main field using the drop-down menu. It is best to select the field based on which the categories are formed
• Specify what you are looking for. For instance type in Summarize this under question 1.
• Hit "Ask relevance"

Another sample:

• Select the main field using the drop-down menu
• Specify what you are looking for and the format in which you expect to see the results. For instance What are the top 3 items in this cluster?
• Hit "Ask relevance"

When the process is finalised a new section appears on the page with information about the data included in the selected category.

Tips

📘

You should think of yourself as the driver - be helpful to the model and steer it where you want by giving it specific keywords and instructions

For good answers

• Give it a specific format and a clear action verb (i.e. prefix with In bullet-point format, list / Summarise concisely in a paragraph)
• Do not ask for x number of sentences - it does what it feels like
• Asking about key sentiments, key statistics if any, what SPECIFIC improvements …
• Be specific about what you want. You can ask what it thinks users want etc., but it will go off on a tangent sometimes or rephrase what is included in the category.
• Main takeaway is that you should think of yourself as the driver - be helpful to the model and steer it where you want by giving it specific keywords and instructions