On this page, we will explain more advanced options when working with clusters on the Explorer app. First, you need to activate the cluster explorer mode as explained on Data and Cluster explorer or as instructed in the image below:
Scroll down to see the individual clusters (i.e. cluster cards) and you will have access to advanced setup options on the bottom right.
There are various ways to label clusters, the three most common ones are 1) type in a label, 2) auto-labelling using aggregation on the most common values of fields in a cluster and 3) auto-labelling using AI.
- To type in the label simply go to the cluster card, click on the edit button, type the label in the edit window and hit save.
- Auto-labelling using an aggregation on the most common values of fields in a cluster is set up when you define what cluster to use. It is a 5 step process as shown in the image below.
- Select the cluster
- Switch to "Custom template" under "Default label rule"
- Type in the '>' symbol and the field name you wish to aggregate on. Matching field name(s) will show up in blue. Hit enter on the field name.
- Auto-labelling using AI
Auto-labelling using an AI model is set up when you define what cluster to use. It is a 4 step process as shown in the image below.
- Select the cluster
- Switch to "AI-generated labels" under "How to label clusters"
- Select the field based on which labels should be generated.
- Apply changes.
In different scenarios, you might want to hide a cluster. For instance, a cluster might not be useful for the insight you are looking for. Simply click on the advanced setup option on the bottom right of the corresponding cluster card and hit "Filter out cluster".
Another way of hiding clusters is to enter the cluster id under the filtering option when clusters are set. You can also use this option to get the clusters back by deselecting the ones in the list.
Warning: This is not undoable. To get the original clusters back, you will need to rerun clustering.
If the content of some clusters is very similar, you might want to merge them together. For instance, when analysing email text, you are likely to have clusters on different ways of starting the email (e.g. Hello, Hi, Dear ..., etc.). It is possible to merge them together into one cluster as general greetings.
To do so, on the bottom right of a cluster card, click on the advance setup, and select "Merge with clusters".
And select the cluster you wish to merge the current cluster with from the drop-down menu.
Updated about 1 month ago