Modern BI is all about helping more people easily get more insights from data. This not only requires making analysis and visualization of traditional numerical data easier – it also requires making other types of data accessible to more people for analysis as well. For example, consider trying to understand the key topics of a collection of news articles. This type of activity (be it news, product reviews, or tweets) is a fundamental activity for wide range of business users, but it is rarely supported in business intelligence solutions.
Today, we’re excited to release 3 new custom visuals for Power BI. The strippets browser, cluster map, and facet key are designed to help you create Power BI solutions for browsing, understanding, analyzing large collections of text.
Three custom visuals for “Power Reading” in Power BI: Strippets Browser, Cluster Map, and Facet Key
You can download these custom visuals from the Visuals Gallery.
These visuals operate on any kind of tagged document content being visualized in Power BI. This can include key phrases, real-world entities such as people, places, and organizations, or abstract concepts that have been tagged in your data. These visuals make a fantastic pairing with the output of text analysis from the Cortana Analytics suite, but they can also be used to help make basic text content an integrated part of your report.
Here’s an example that illustrates their use for news reading, using the likely familiar example of tracking the “hottest” topics in technology.
(expand the report into full screen view for the best experience)
As you can see in this example, by using these visuals in Power BI we’ve made understanding a large feed of news articles quick and highly visual. Let’s take a look at the three new custom visuals used to make this report.
Strippets Browser: A quick way to view document contents
The Strippets Browser is a document reader that provides two complementary ways of sampling the contents of a collection of documents or a news stream. The first unwraps each story into a thumbnail that reveals the source, headline, and leading text and image of the story. Using the thumbnails view, the reader can scroll through a list or grid or thumbnails to quickly grasp the topic of each story in the collection. Clicking on a thumbnail activates a reading mode in which the full document contents can be read in context. Here is an example from the several hundred most recent documents returned by the Bing query “Microsoft AI”:
The Strippets Browser thumbnails view uses content-based text summaries to guide document reading
Notice the highlighting in the document text? This indicates a reference to a named entity, for example a person, place, or organization. In this example, that entity is the organization “Microsoft”. The first reference to any entity that plays a key role in the story is also accompanied by an icon in an outline strip next to the document scrollbar. Icons are positioned within this strip according to the document locations where the corresponding entities are introduced, and also act as hyperlinks that can automatically scroll the document to these locations. This outline strip, or “strippet”, provides a compact visual summary of the key entities participating in the story as well as an aid to navigation. These strips are also the foundation of the second Strippets Browser mode: the outlines view.
The outlines view collapses each document down to its outline strip and packs these strips together to create visual cross-section of a document collection. If the thumbnails view is like partially unwrapping news stories to take a peek, the outlines view is like slicing them in half to have a good look inside:
The Strippets Browser thumbnails view uses entity-based visual summaries to guide document reading
Both views of Strippets Browser use infinite scroll to continuously retrieve new stories on demand.
Cluster Map: visually grouping similar documents
Even with the reading efficiency of Strippets Browser, sometimes you might just want to focus on a particular topic, or see the range of topics in a document collection before deciding where to start. Enter the Cluster Map, which displays clusters of related documents in an expressive image mosaic. Clusters can be arranged in a symmetric “spiral” layout or a more freeform “relational” layout, with cluster proximity in the latter case determined by the relatedness between clusters. Selecting a cluster in either layout can filter and highlight stories in Strippets Browser when both are used in the same report:
Cluster Map filtering of Strippets Browser. Left: spiral layout. Right: relational layout with cluster selected
The Cluster Map is a great way to help users navigate a report by providing a visual way to select everything related to a certain topic and quickly browse a large variety of content.
Facet Key: Filtering documents by shared attributes
The Facet Key is the final piece for putting together a great document analysis report, showing the most frequent entities of various types across an entire document collection. Selecting an entity of interest filters and highlights linked visuals in ways that enable document collections to be analyzed systematically, one entity at a time.
When a Facet Key is used in the same report as a Cluster Map, entity selections trigger blue highlight arcs around clusters representing the proportion of clustered documents containing references to that entity. Clusters containing no references are also faded out for contrast:
Facet Key highlighting of Cluster Map. Blue arcs show document matches for person “Satya Nadella”
When a Facet Key is used with a Strippets Browser, entity selection triggers blue highlighting of the corresponding icon across all matching strippets, filtering out other strippets automatically:
Facet Key highlighting of Strippets Browser. Blue marks show matches for organization “Microsoft”
An additional benefit of the Facet Key is that it can be used to create and interpret icon mappings in other linked visuals. In this example, the top three entities of each type are assigned a color, where hue represents type (e.g., red locations versus orange organizations) and saturation conveys the rank of that entity (more colorful means more frequent). Icons can also be customized using Font Awesome CSS class names for even more visual awesomeness.
Try out these newsworthy Power BI custom visuals
All three custom visuals can be used independently or in conjunction with other Power BI visuals. They are designed to be better together, and for structured reading scenarios like the entity-based news reading example described in this blog, but we’re excited to see how you use them to unlock new insights on new types of data!