Explorer: Category / Cluster View
Present insights and documents within a dataset under configurable categories
Under various circumstances, data can be grouped into different categories. Categories might:
- Already exist in the dataset (e.g. months, department names, age groups, etc.)
- Be an outcome of a clustering workflow
On this page, we are learning the basic steps for setting up a category view on Explore dashboard.
Add a Category View tab/page
To add a category view to an Explorer dashboard click on the
+ sign on the top bar, and select "Category View" from the drop-down menu.
Type your desired name for the view, select a color for the tab/page (optional) and hit add page.
Select the main field based on which the dataset is categorised
On the new page that pops open automatically, click on the "Select Category" button. Next, you can choose either of the following options:
- Select an AI-generated clustering result. Note that to use AI for clustering or tagging, you need to follow our guide on AI-clustering or AI-tagging workflows.
- Any of the categorical fields that exist in your dataset
When done, you can see a new tab/page is added to the top bar.
Note: Categorical fields contain a finite list of values that are repeated among the items in a dataset. For instance, in a dataset of 1000 customer queries, if there is a field identifying the month the query came in, the values in this field are 12 month names repeated among the 1000 query documents; therefore, month is a categorical value. Same applies to fields such as gender, names of departments within an organization, country of origin, etc.
In this section, we introduce the main components present on a category view page and the corresponding configuration. More advanced components are introduced on their own separate documentation pages.
The overview chart
After selecting categories, the dataset is broken and presented based on the chosen field. On the top, a chart illustrates the categories/clusters with their statistics. Using the menus on the right, you can indicate how much of the data should be presented on the chart (e.g. all, summary), representation type (e.g. bar chart, column chart, etc.), as well as the metric based on which the data is shown and sorted.
Note: By default frequency is used to sort the data. More items appear on the list after defining metrics to apply to the dataset (e.g. average price, minimum ranking, sum of the grades).
Add a metric
Relevance AI provides you with common metric analysis (i.e. analysis such as average, minimum, maximum, mean, etc. that are applicable to numeric values). You can view the results as well as sort the categories according to metrics.
To set up a metrics, click on "Metrics" on top.
Click on "Add metric to analyse"
From the two drop down menus: select the field to which you wish to apply the analysis, and type of analysis (e.g. avg, min, max, etc). Click on "Continue" to finalize the setup.
You can see the new metrics is added to the metrics menu on top and can be used to visualise data.
Note: You can define as many metrics as you wish following the process explained above.
Further down on the page, you can see single categories/clusters that are listed after each other. For each cluster/category, the top bar presents number of items in each category, along side with the result of the defined metrics analysis applied to documents within the category.
Tip: a shortcut to the metrics setup page exist on the right-hand side of each cluster bar.
The image below explains some of the key features on the page.
Set up document templates
To set up what fields to see in each document card, click on "Template".
A window will open in which you can select the primary (image and text) field(s), the secondary text field, and up to 6 metadata fields which will be shown in smaller font on the main page.
x on the right side of selected fields to remove them from the view. You can reorder the selected fields by drag and drop on the icon on left side of selected each field.
Aggregation and charts
The dashboard allows you to set up charts and aggregations (i.e. grouping) within each cluster. Simply click on "Chart" and then "Add chart" as shown in the image below.
Type in a name for your chart, set the metrics by clicking on "Metrics to analyse", configure the grouping by clicking on "Add category to group by" and select your desired type of chart.
A sample setup is shown in the image below where
Stars is the chart name, average on the
stars field is calculated, the results are grouped based on the
language field and shown as a
The result is placed on the right side of document cards
Note1: Multiple charts can be added by clicking on add on top
Note2: Multiple metrics and grouping can be added to each chart by using "Add metrics" and "Add category".
Note3: Remove the designed charts by going to the setup page and clicking on "Remove chart" at the bottom of the page.
Note4: General charts supported by Relevance AI are bar and column charts, word and bubble charts and time series. There are also charts for specific analysis such as sentiment where general overview or the trend over time (time series) chart can be used. More chart specific features and configurations are explained on Chart view
Delete a tab/page
Simply click on the down arrow on the right side of the tab, hit "Delete page" and confirm the action.
Updated 3 months ago