On-site menu
Renamed "Analytics" to "Daylight"
Option for generating support logs
SAML improvements:
Option to disable local password authentication when SAML is enabled
Option to enable or disable SAML Request Signing
Option to enable or disable SAML Assertion Encryption
Option to not set the SP Cert and Private Key
Native AWS Amazon Machine Image (AMI) support
AMI On-site image is available, which can be natively deployed in an AWS environment
On-site images continue to be available in other formats: VMWare, KVM, and VirtualBox
Automatic restart
When the system starts, Daylight will automatically start, requiring no manual stop/starts during system reboot
Highlights
Sentiment
Create projects with Polish documents
“Concept Cloud” is now “Galaxy”
“Compare Concepts” is now “Volume”
“Score Drivers” is now “Drivers”
Projects List: Account selection is retained
Filter Pane: All filter options are collapsed by default
Concept Details Pane: When the concept search term changes, the scroll position is reset
Concept Details Pane: Delete a saved concept
Concept Details Pane: Downloading matching documents includes each document’s sentiment
Volume: Sort order name “Most relevant” has changed to “Most prevalent”
Volume and Sentiment: Concept lists show sortable fields and current sort order
Drivers: Scatterplot y-axis is now Average Score
Drivers: Feature configuration now located in the left pane
A new version of the API client is now released to PyPI
User logins are no longer case sensitive
Upload Documents: If a document contains a bad metadata field, the error message will include the specific field as well as the document causing the problem
Adding or removing a concept from a saved compound concept no longer grays out the Concept Detail Pane’s Match Counts, Top Related Concepts, and Matching Documents sections
An informational message now displays when project filters match zero documents
Volume: When you refresh the page, you’ll still see your selected concept on the Concept List
Drivers
In the Excel download, a saved compound concept will now reflect the name you gave it, as opposed to the words or phrases that it contains
Adjusting the filter now retains the selected concept
Projects of significant size will now show results when the concept selection is changed from suggested drivers to saved concepts
Text labels for points in the scatterplot will now stay within the visualization area
When viewing Saved Concepts, a newly saved concept is added in real-time
Search-as-you-type now properly filters results
Highlights is the new project home page, and features a new name and icon. It offers an orientation to what each Daylight feature can offer.
Surface key project details such as project language and metadata types
Manage project attributes, such as sharing and copying URLs, and preview insights from each feature
Simple visualizations of the capabilities of each feature
Navigate directly into each feature to access concept selection, filtering, and sorting
Sentiment is a new Daylight feature that provides document level sentiment for concepts, identifying the words and phrases that customers feel most strongly about. It is reflected by its new name and icon.
Immediately view a project’s overall sentiment distribution
Understand the polarity of each concept’s sentiment (positive, neutral, and negative)
Surface the top 50 words and phrases with the most sentiment
Filter subsets of documents to examine differences in sentiment
Native language support for 15 languages is included: Arabic, Indonesian, Chinese, Dutch, English, French, German, Italian, Japanese, Korean, Polish, Portuguese, Russian, Spanish, and Swedish
Emojis, emoticons, and compound concepts are analyzed
Luminoso now natively supports documents in the Polish language. "Świetnie!" You will now see Polish as an additional option in the language dropdown in the “create a new project” section.
Concept Cloud is now Galaxy. While the feature has not changed, all references to it will reflect its new name and icon.
A word cloud is a two-dimensional visualization of text data, where the size of each word it contains indicates its frequency in the dataset. Words are placed next to each other as it is generated for aesthetic, rather than functional, effect.
The Galaxy feature provides much more than a traditional word cloud. It’s comprised of 150 dimensions of associations between concepts. Word cloud frequency can mislead, as there are words which are common in basic expression that don’t actually confer analytic value. Instead, Galaxy prioritizes the concepts that are more prevalent in your data set than would be expected to appear in the underlying language. Each concept is alongside its most closely associated concepts, and seven of the largest clusters of conversation are highlighted in different colors for at-a-glance insight.
Try it: click and drag any concept, and as you move your cursor, the Galaxy spins to help you visualize its 150 dimensions. Letting go of the concept flattens the view as if it were under a pane of glass, so you can perform quadrant analysis in two dimensions.
Volume, formerly known as Compare Concepts, identifies the relative volume of individual concepts contained in the project dataset, with and without breakdowns by metadata field. The feature has been renamed to reflect its core analyses and value.
The feature name change (and updated icon) is now reflected in the navigation menu.
While the feature has not changed, all references to it will reflect its new name. There is no change to the get concept/score_driver API endpoint.
Account selection on the Projects List page is retained when refreshing the page, after deleting a project, or clicking the “Return to projects list” link.
Filter options are now collapsed by default in the filter panel, to make it easier to hone in on a desired filter option. Once a filter option is in use, it will remain expanded to provide visual confirmation of the current filter’s state.
When the concept search term changes, the scroll position is reset to the top of the Top Related Concepts and Matching Documents sections.
You can now delete a saved concept through the Concept Detail Pane in any Daylight feature.
Each document’s sentiment (Positive, Negative, or Neutral) is now provided when downloading matching documents from the Concept Detail pane.
The Volume and Sentiment concept lists now designate any column that is sortable with a double grayed-out arrow. The current sort order is shown via an up (ascending) or down (descending) arrow. To change the sort order, simply click on one of the sortable column’s headers.
The Drivers feature in Daylight understands words and phrases and correlates them to scores, so teams can quickly implement fixes and improvements. Now, the y-axis of the scatterplot provides a clearer picture of what’s happening: concepts are plotted at their individual average scores, versus the dashed line representing the overall average score of the dataset.
Concepts with a higher average score than the overall average score are in blue. Concepts with a lower average score than the overall average score are in red. Selecting or hovering over a concept provides an updated tooltip, displaying the concept’s difference from the overall average score.
To provide greater focus on the scatterplot visualization, the Drivers feature configuration options are now located in the left pane below the Filter options.
Please use the command pip install luminoso-api to access the latest version. The command from luminoso_api import LuminosoClient will provide the v5 API by default. You can still explicitly request the client for either API version using from luminoso_api.v4_client import LuminosoClient or from luminoso_api.v5_client import LuminosoClient. The new version will automatically pull forward your old API tokens when connecting to daylight.luminoso.com.
String fields with many values are displayed as search-as-you-type filters. In the example below, typing “I” within the Brew Name field brings up the desired value “IPA”. Clicking on that “IPA” value will now properly filter the Drivers results, rather than displaying an error.