Note: This article contains references to a previous version of the Stratifyd interface.
Chapter 7.1 Product Release Notes
Stratifyd is excited to share with you our Chapter 7.1 release as a follow up to our Chapter 7.0 release – Fusion, Insights, and Action. In Chapter 7.1, the Stratifyd platform story continues with a focus on enhanced topic modeling with a new Neural Topic Model; enhancements to Focused Insights introduced in 7.0; an optimized taxonomy building and editing process through a new Taxonomy Editor; and improved team productivity through Team Templates and a new platform onboarding process. The Chapter 7.1 release roll out will begin on November 9th, 2020.
Neural Topic Model (NTM)
A new model to advance the relevancy, richness, and meaningfulness of new topic discovery.
In the Chapter 7.1 release, Stratifyd is excited to bring you a new Neural Topic Model (NTM) which is an unsupervised natural language processing model for analyzing unstructured textual data. This model is the next evolution in topic modeling, employing the most cutting-edge NLP and machine learning techniques to discover interpretable underlying abstract "topics" and organize the documents accordingly. NTM will be available as an additional model in the Stratifyd platform and will produce similar field/widget outcomes as Auto-Topic Predictive, but with greater accuracy and depth.
Unstructured data can offer a gold mine of customer and market insights, but its inherent messiness can make it a challenge to uncover the insights that matter.Too often traditional topic modeling leaves insights on the table with ambiguous and inconsequential topics. The new Neural Topic Model advances your ability to uncover rich, meaningful, and relevant topics in unstructured data by applying the latest in neural networks resulting in greater value from your unstructured data.
The NTM offers key benefits beyond traditional topic modeling including:
• Produces higher quality topics through cutting-edge neural network structures that lead to ~10-20% better topic coherence and uniqueness.
• Offers greater topic richness through topic outputs of up to 5-word phrases vs traditional bigrams (2-word phrases).
• Offers improved ability to capture meaningful results and semantics with both short and long data records/verbatims.
• Provides faster and more stable training through a pre-trained language model that is not dependent on lexicon dictionary and semi-supervised customization allowing you to enhance NTM through training data set labelling.
How to Implement
Within the platform, both Auto-Topic Predictive and Neural Topic Modeling will be available as model selection options. Both models will offer similar field/widget options for topic and sentiment reporting and visualizations. Currently, the NTM is only available in English with additional languages to come. Please reach out to your Customer Success Manger or account team to talk more about applying the NTM to your program.
Focused Insights Updates
Identifying emerging topics in the data set now with rolling time filters and a summary view.
In the Chapter 7 release, Stratifyd created the ability to identify emerging or “fresh topics” in the data set over specified time frames to take the work out of new topic discovery. In the 7.1 release, we’ve continued to enhance this feature with three key updates:
Rolling time filters have been enhanced to allow for custom ranges and calculation of time ranges.
A topic summarization has been included to provide greater clarity around the primary theme of the emerging topic.
As a reminder, Stratifyd has included Focused Insights because we understand the pain of identifying new themes in your data over time. Focused Insights will automate the process of surfacing emerging topics to help users quickly identify new topics that would be difficult to find via traditional analysis becoming your virtual, 24/7 analytics partner identifying what’s new and emerging in your topic data.
Optimize your taxonomy more efficiently and effectively through a taxonomy preview editor.
Stratifyd is excited to introduce a taxonomy editor interface which provides the ability to modify the rules in a taxonomy topic and then run a "draft" version of the updated topic(s) to reveal a preview of how many new verbatims were added and how many were removed. The taxonomy editor also now provides examples of those verbatims for review. This preview will allow users to preview the changes in their taxonomy without the burden of fully reprocessing the data set.
Stratifyd knows that building and optimizing a high-quality taxonomy requires a significant investment of time and effort. Previously, modifying a taxonomy required saving changes, reprocessing of the data, waiting for the results to be populated in the dashboard, and digging in the dashboard to determine how the data changed as a result of the taxonomy modifications. With the new taxonomy editor, taxonomy modifications not only can be made more efficiently, but it is significantly easier to gauge the effectiveness of those modifications. The result is a faster path to a higher quality taxonomy that leads to high quality insights.
How to Implement
To use this feature, when editing your taxonomy, click the “preview taxonomy” button at the top right of the page. The taxonomy preview window will appear. The results of the preview will default to processing 1000 records with options to include more records and will provide insights into labelled documents. matched rules, and examples of the verbatims. From the editor, new logic can easily be added to existing taxonomy rules.
Simplifying the process of accessing and sharing dashboard team templates.
Now it is easier than ever to access or share a dashboard template with the new "Team Template" category under the left-hand navigation of the dashboard view. Users will see templates that have been shared with them and can edit a template from this area.
This new navigation simplifies the process of building impactful dashboards and supports organizational dashboard governance. The Team Template category also helps to focus new platform users on their available templates as a tool to help jump start their time-to-insights.
New User Onboarding
Simplifying the onboarding process for first time users with a new registration process and redirect to shared assets.
In 7.1, we’re working to simplify the onboarding process for new users by:
Enhancing the registration process by gathering additional user detail to help tailor the future user experience.
Directing users immediately after the registration process to their library of assets or templates that have been shared with them.
This streamlined experience helps more quickly orient and focus new users on the assets that are valuable to them.