Release Notes
Keep up-to-date with new features and improvements
The NovaceneAI Platform is continuously improving. New releases are delivered every three weeks, on Saturday mornings between 9am-12pm EST.
Next Release
- November 9th, 2024: Release 2.11.1
Latest Release Notes
October 19, 2024: Release 2.10.1
Summary: Faster, better, and more open ends. Now, with the new Multi-column Clause Extraction enricher, text from multiple columns is automatically merged so that you can analyze all the text in your dataset at once. This release includes changes that make the Stage more responsive—dashboard widgets now update faster than ever when filtering and asking questions about your data.
Multi-Column Clause Extraction
If your dataset contains multiple columns of text that need to be analyzed, the new Multi-Column Clause Extraction enricher will merge selected columns into a single column. From there, you can run enrichments on your entire dataset at once, giving you a more holistic view of the topics in your dataset.
To use this enricher, select Multi-Column Clause Extraction select a column with text you’d like to analyze, and run Multi-Column Clause Extraction. Once it has completed, open the enriched file, repeat the process on the next column until every column of interest has been run.
At the end of this process, all of your open-ended questions will be merged into two columns at the far right of your dataset. The column named “Column Name” will contain the text from the headers of your selected columns, and the column named “Clause” will contain the clause-extracted text from your columns. From there, you can run any other enrichments you need on your clauses, including Autopilot.
Stage – Responsivity of Widgets
This release includes improvements to the speed of loading and interacting with elements on the stage. The amount of time it takes to add and remove filters has been reduced by ~20 seconds, and the amount of time it takes to load reports has been reduced.
Stage – Show [Blank] Values
Previously, empty cells or blank values were not displayed in the stage. Now, they are. This can help you understand the proportion of different values in your data better. Blank cells in a column will now be displayed as [Blank] in visualizations in the stage. You can now also filter by [Blank] values.
Clustering-Dependent Enrichers – Issue Resolved
Some users reported issues when using the Cluster Label and LLM Cluster Theme Extraction enrichers. These issues have been resolved. Please note that you need to run clustering on your data prior to running these enrichers, as they depend on information created in the clustering process.
September 28, 2024: Release 2.9.2
Summary: Release 2.9.2 includes four new enrichers that allow you to cluster and classify your data in new ways. Custom Theme Classification allows you to choose the themes you’d like to quantify in your dataset. Thematic Classification works particularly well on larger datasets, and allows you to set a maximum number of themes. Sentiment Analysis GPT detects sentiment, emotion, and whether a piece of text is sarcastic or sincere. The new Offline Language Translator allows on-prem and air gapped clients to do translation offline.
New Enricher – Custom Theme Classification
Have you ever wanted to classify your data into specific categories? Are there specific themes you need to quantify? Now you can!
The new Custom Theme Classification enricher uses an LLM to sort your text data into categories of your choosing. By default, GPT 4o Mini is used for this enricher. Please contact support if you would like to change your environment’s LLM selection. When you select Custom Theme Classification from the enrichment dropdown, you can enter the different themes you’d like to quantify in your data. You can add or remove themes with “Add Theme” and delete buttons.
For best results, either run Autopilot or Clause Extraction on your dataset prior to using Custom Theme Classification. Target the column with your clause extracted text when running this enricher.
New Enricher – Thematic Classification
This new classification enricher is designed for large datasets. When using Thematic Classification, you can choose the maximum amount of themes you’d like to receive. You can also choose to merge similar themes with a toggle when running this enricher.
For best results, either run Autopilot or Clause Extraction on your dataset prior to using Thematic Classification. Target the column with your clause extracted text when running this enricher.
New Enricher – Sentiment Analysis GPT
This new LLM-powered enricher adds three new columns to your data to help you understand sentiment, emotion, and detect sarcasm—all at once. By default, GPT 4o Mini is used for this enricher. Please contact support if you would like to change your environment’s LLM selection.
New Enricher – Language Translator (Offline)
A new offline language translator enricher is available through the front end and over API. This translator is designed for clients that run the NovaceneAI platform in an air gapped or on-prem environment. Similarly to the online language translator, this enricher translates all text in the selected column to English.
The offline language translator supports the following 25 languages:
Arabic | Dutch | Greek | Italian | Russian |
Catalan | Esperanto | Hebrew | Japanese | Spanish |
Chinese | Finnish | Hindi | Korean | Swedish |
Czech | French | Hungarian | Persian | Turkish |
Danish | German | Indonesian | Portuguese | Ukrainian |
Studio – Unique URLS for Datasets
Each file viewed in the Studio now has a unique URL. You can use this URL to share specific datasets with other users that with access to the dataset’s project folder.
Studio – Train Models Control Moved
Previously in the Studio, users could train models on a given dataset under the “Train” tab of the inspector bar. This function has been moved to the Models screen, and has now been removed from the Studio screen.
Stage – Abrupt Re-render Issue Resolved
Some users experienced an abrupt re-rendering of the default dashboard when loading a custom report. This has been resolved.
Archive
- September 7, 2024: Release 2.9.1: Includes the new ability to navigate to a specific dataset with a unique URL. This release also includes a number of resolutions for issues in the Stage. More…