Analytics engine output
The Stratifyd analytics engine generates new data points based on your unstructured data. Here are the ways in which we represent these data points.
The analytics engine reads through every piece of text in a dataset and compares every possible combination of words to find statistically significant n-grams.
An n-gram is a bi-gram by default, but since you have the option to change the n-gram length in the advanced settings of the Unsupervised NLU model and the N-Gram analysis, we call it an n-gram.
Buzzwords constitute a compilation of every statistically significant n-gram found in your textual data. In order for an n-gram to be deemed statistically significant, we look at how often the words occur together in the dataset versus how often they occur individually.
With a buzzword cloud visualization, the Unsupervised NLU model displays n-grams in a cloud of words in various sizes and colors.
- The font size represents the statistical significance of the n-gram.
- The color represents the sentiment associated with the n-gram.
With a buzzword list visualization, the Unsupervised NLU model displays the n-grams in a list along with other columns of values.
- The n-grams are listed in order of highest occurrence to lowest.
- The second column displays the occurrence count for each n-gram.
- The third column displays the sentiment value for each n-gram.
- If you mapped a date field, a fourth column displays a sparkline showing how the values change over time.
You can add this list information to a topic wheel Pie chart by turning on the Key N-Grams option.
To generate semantic topics, the Unsupervised NLU model performs unsupervised machine learning on top of the n-grams to determine hidden themes, and then groups the documents accordingly.
A document may occur in multiple topics, so if you add up the percentages of documents contained in each topic, the sum is greater than 100%.
You can display topics in a pie visualization or a treemap visualization.
With a Pie visualization, each topic is represented by a slice of the pie, beginning with the largest at the 12:00 position and moving in a clockwise direction.
- The words around each slice represent the top one or two n-grams for the topic.
- The order and size is determined by the statistical relevance or “tightness” of the topic.
- The color represents the sentiment of the topic.
With a Treemap visualization, each topic is represented by a rectangle, beginning with the largest at the bottom right and moving up and to the left.
- The words in each rectangle represent the top n-grams for the topic.
- The order and size is determined by the statistical relevance of the topic.
- The color represents the sentiment of the topic.
Adding a time element to your structured and unstructured data helps you to understand trends and patterns and to discover relationships across different data sets.
In order to use temporal data with Unsupervised NLU model analyses, you must map a Date field when you set up data mappings.
With a chart visualization, you can create a timeline showing topic trends where each bar represents a topic while the bar groups represent a time period.
- The date field from the Unsupervised NLU model is the X axis.
- The topics field from the Unsupervised NLU model is grouped.
- The Number of Records field from Calculated Fields is the Y axis. CHECK THIS
With a pie visualization, you can turn on the Temporal Trends option to display the same data below a topic wheel using only the topics field from the Unsupervised NLU model.
- The date automatically becomes the X axis.
- The Style is set to Bar to show grouped topics for each date.
- The number of records automatically becomes the Y axis.
Interactions and filters
Every time you click a data element in the dashboard, it applies a filter on your data. The filter appears to the top right corner of the page next to Add Widget button.
Click the filter icon to open a dialog where you can change it, delete it, add other values, or select how to handle records with no value in this field.
Filters do not persist when you leave the dashboard unless you save them as segments. See Filtering and segmenting data for more information.
You can set filters at three levels.
To set a filter on the every widget and every tab on your dashboard, click the filter icon in top right corner of your dashboard or in a side panel.
This opens the Global filter workspace for your dashboard where you can add simple filters.
Click Advanced to open a workspace where you can visually combine Boolean filters.
To set a filter on only a single tab of your dashboard, click the tab and select Filter.
In the Tab filter workspace, as with the Global filter workspace, you can add filters, use the advanced version to create Boolean filters, and save filters as segments.
In the top right corner, the Override Global Filter applies the filter selections you make and freezes the visualizations so that the tab is no longer interactive and selections on other tabs do not affect the data on this tab. Leave the override checkbox cleared if you want to interact with the data and allow interactions with other tabs to affect the data on this tab.
When you apply tab-level filters, the filter icon appears on the tab so that anyone using the dashboard can see that it is filtered.
If you opt to override global filters, the filter icon appears in red.
To open the Widget filter workspace, mouse over the widget to reveal the menu button at the top right, click it and select Filters.
As with tab-level filters, the Override Global Filter applies the filter selections you make and freezes the visualization so that the widget is no longer interactive and selections on other widgets do not affect the data on this widget. Leave the override checkbox cleared if you want to interact with the data and allow interactions with other widgets to affect the data on this widget.