When dealing with a big amount of data, clustering helps to break the data into meaningful categories. A useful next step is to label each category. Relevance AI provides you with the ability to
- manually label clusters, also edit the labels
- automatically generate a label for each cluster via AI
This is possible on the Explorer dashboard and under the category/cluster view. So make sure a category/cluster view tab/page is setup with AI generated clusters.
Select an AI generated clustering result on a category/cluster view tab/page. A table similar to the image below is shown on top. You will notice that clusters are labelled with their ids (e.g. cluster-1, cluster-2, etc.)
There are a few ways to access the manual labelling:
- Click on the bar on the top table. A page will open including all the entries in the cluster. Click on the pencil icon on top.
- Scroll down on the page to reach the cluster whose label you wish to modify and click on the pencil icon on top.
Under both ways above a small window similar to what is shown on the image below will open. Enter the desired label for the cluster and hit "confirm".
Note that the changes are reflected both on the tables on the top view, and down on the page when each cluster data is presented.
Auto labelling clusters is when we use AI to label clusters based on the concepts presented in each one. The Explorer dashboard provides you with an easy access to state of the art AI models that generate label for your configured dashboard. To be able to use it please follow the instruction below.
- Manually label 3, 4 clusters. This is to guide the model on your preferred writing style. Note that there is no need to start from the first cluster or continue in order. Any clusters can be chosen for manual labelling.
- Locate "Auto label AI categories" after the top section on the right side. This is to label all clusters at once.
When "Auto label AI categories" is clicked, a window will open on which you will select the field based on which the labelling must be done, this field is usually the same one used for clustering. And click on the "AI auto label categories" button.
The process is rather quick, but the execution time varies depending on the number of clusters. When the task is finalised, all cluster ids will be replaced with AI generated labels.
Note1: Text generation is one of the most complex tasks in AI. Therefore, even under the state of the art models, human modification as a post-processing step might be required.
Note2: If not enough clusters are manually labelled before hitting "Auto label AI categories", an error message will open asking you to do so.
Note3: You can auto-label clusters individually . To do so follow the above mentioned instruction for manual labelling but instead of typing in the label, click on "AI auto-label". The rest of the process is as what was explained above.
There is another practice similar to labelling but unlike clustering it is to perform one-to-many categorisation. You can read more about them on Tags.
Updated 20 days ago