Reveal AI Release Notes V3.00 GA
Release Notes – Reveal AI Version 3.00 GA (Release Date: 02/28/2022)
Reveal AI 3.00. continues Reveal’s committment to “all in on AI“ with an industry-leading upgrade to supervised machine learning and AI library models. The flagship feature of the 3.00 release is High Precision Classification (HPC). A high-level overview of it is also available at High Precision Classification under COSMIC (Cognitive Machine Learning).
Type | Category | Item | Description |
|---|---|---|---|
New | Supervised Machine Learning | High Precision Classification (HPC) | HPC allows for a greater level of interaction and guidance between the user and the AI engine. The model is trained on cleaner input and less noise and identifies most relevant passages that the user can see as highlights. In addition to providing a tag to a document under review, the user highlights the portion of the document that makes it responsive. This way, the model receives only the information relevant to the case as training data. As the model scores each document, the user will be able to see the highlights of the most relevant passages of the documents identified by the model. For a user guide of HPC, consult the documentation at High Precision Classification. |
New | Natural Language Processing | Roundabout Writing Style | A new emotional intelligence score as part of pre-run analytics on cases in Reveal AI. This score is based on stylometric analysis of written content and intended to detect conversations that are unusually wordy and avoid addressing the point directly. Learn more about Roundabout Style at https://resource.revealdata.com/en/blog/stylometry-and-the-fraud-triangle. |
New | Natural Language Processing | Emotional Score Normalization | Sentiment and other scores already available in Reveal AI will now be normalized to be between the values 0 and 10 for simplified searching and relative comparison. |
New | Supervised Machine Learning | Score Report | Add ability to open a document-level score report from the document viewer to show one or more of the scores assigned by the COSMIC AI engine to the whole document, individual segments, High precision highlights, and expand to see scores for specific features within the document. |
New | Model Library | Business Meetings | The Business Meetings model identifies communications that discuss or mention business meetings—whether they occurred in the past, are currently ongoing, or are planned for the future. That includes recent meetings in collaboration platforms such as Teams, Skype, and Zoom. The model is trained to also look for conversation that imply a meeting, even if one is not explicitly mentioned. |
New | Model Library | MBERT Models | Multilingual BERT models offer contextual language understanding in over 100 languages. MBERT starter models from Reveal AI Model Library can be applied as any other library models. A number of models in the Reveal AI Model Library have MBERT support, including
|
Update | Insights Dashboard | Module Titles | Allow custom naming (titles) for modules added to the Insights dashboard. |
Update | User Experience | Term Reports | Updated the display of error message to include more details when validating Regular Expression syntax in Term Reports. |
Update | User Experience | Storybooks | Enable selecting a group to have access to a Storybook during the creation process so save clicks. |
Update | User Experience | Build Entity Model | Provide clearer error message when building of an entity model fails. |
Update | Administration | Users | Rename the Delete user option to Archive and display the email of the user being archived on the confirmation screen. |
Update | Administration | Storybook Creation | Increase the maximum number of returned storybook database server connections from 50 to 500 when creating new storybooks. |
Update | Logging | File Manager | Include the path in the error log produced by OnPremiseFileManager when a file or folder cannot be accessed. |
Update | Data Connections | Relativity® | Restores SSO login from RelativityOne to Reveal AI. This SSO option requires configuration under Tenant Setup for ClientId, ClientSecret, RelativityUrl. The documentation showing these new settings is available. |
Fix | User Experience | Saved Searches | Corrected dropdown menu behavior to allow long lists of saved searches to display properly |
Fix | Administration | SQL Server | Fixed the creation of SQL server connection for the default tenant. |
Fix | Administration | Export | Fix ability to edit an existing export record in System Admin area. |
Fix | Administration | Export | Fixed the Save & Run Now button for exporting data from a storybook. |
Fix | Connectors | Relativity® | Corrected error when attempting to create a field with COSMIC groups in Relativity® |
Fix | Threading | Inclusive Documents | Addressed a specific issue with documents in Japanese where segment merging occurred more often than it should with MergeReason #8. |
Fix | Threading | Combining Segments | Addressed a specific issue primarily with documents in Japanese where combining segments results in inability to render segments in the Document viewer. |
Known issue in RelativityOne hotfix in Chrome/Edge. The SSO redirect worked as tested in February 2022 in Firefox, Chrome, and Edge. RelativityOne has released a hotfix within the last week that introduced an infinite redirect message with Chrome and Edge. Firefox still works. The issue cannot be corrected on Reveal AI’s side. There is an open ticket with Relativity® and potential fixes to the redirect issue are awaiting updates from RelativityOne.