32.0.28 (14 Dec 2022)
Document Classification
Fixed
Classification of pages as blank– We've fixed an issue that caused pages containing only a set of elements that were positioned closely together to be misclassified as blank.
Box Integration
Fixed
Processing of fields marked as illegible– Previously, if a submission contained fields marked as illegible during Transcription Supervision, a 400 error would occur in Box. This error occurred because values of illegible fields are recorded as null in Hyperscience, and Box does not support null values. A fix for this issue is included in v32.0.28, and Hyperscience ignores illegible fields when sending their submissions to Box.
32.0.27 (4 Nov 2022)
Classification
Fixed
Blank pages and Machine Classification – We’ve fixed an issue that caused blank pages to be misclassified during Machine Classification in some circumstances.
32.0.26 (18 Oct 2022)
Layout Editor
Updated
Deleting multiple fields in the Layout Editor – You can now delete multiple fields in the Layout Editor at the same time.
Making changes to shared fields – We've reduced the time it takes for the system to register changes made to shared fields in the Layout Editor.
Reporting
Updated
Adding new CSV files to the Usage report – We’ve added the following CSV files to the Usage report (Reporting > Usage):
settings.csv – contains all application settings.
db_entity_counts_YYYYMMDD_HHMM.csv – contains counts of database entities.
Fixed
Software Version column in the Usage report – We’ve fixed an issue that prevented the Software Version column in the Usage report (Reporting > Usage) from being populated.
32.0.25 (23 Aug 2022)
Layout Editor
Fixed
Deactivating fields in Structured layout variations – Previously, deactivating fields in Structured layout variations prevented these fields from being displayed in the layout editor’s Inactive items list. If you clicked the Commit changes button after deactivating some fields in a layout variation, the same issue also prevented you from opening this layout variation. A fix for this issue is included in v32.0.25.
32.0.24 (15 Aug 2022)
Integrations
Fixed
Submissions created from the Email Listener Block – If the folder specified in Post Process Archive Folder did not exist, the Email Listener Block created an infinite number of submissions. With the fix included in this version, the Email Listener Block creates submissions only if the folder specified in Post Process Archive Folder exists.
Security
Fixed
HttpResponse usage – To address security vulnerabilities, we've made updates to the system's usage of HttpResponse, including adding missing content-type information and output encoding.
Upgrading okhttp to 4.10.0 – To address a security vulnerability, we've upgraded okhttp to the latest stable version, 4.10.0.
Caching of pages containing submission images – To guard against unauthorized access to sensitive information, the system now prevents pages containing submission images from being cached by intermediary proxy servers and local web browsers.
Security
Fixed
Accessing /admin – We've resolved an issue that caused a 500 error to occur when attempting to access the /admin section of TVE instances.
32.0.23 (19 Jul 2022)
Submission Processing
Fixed
Redirecting users after completing submission tasks – Previously, when completing submission tasks, the system redirected users to the Submissions table. With the fix included in this version, the system now redirects users to a loading screen until a subsequent submission task is available.
System & Health
Fixed
Clicking the “Reset settings to default” button after changing the “Gather health statistics” setting – We’ve fixed an issue that resulted in an unexpected error when clicking the Reset settings to default button after changing the Gather health statistics setting in Settings (Administration > Settings).
Security
Fixed
Authorization for /admin URLs – To address security vulnerabilities, we've updated the permissions required to access several /admin URLs that allow users to upload files or execute code.
32.0.22 (8 Jul 2022)
Supervision
Fixed
Submitting a single task multiple times – We've resolved an issue that caused Supervision tasks to be submitted multiple times and in place of other tasks in certain situations.
All fields pointing to the same bounding box during Flexible Extraction – We've resolved an issue that caused all fields in a document to point to the same bounding box during Flexible Extraction tasks. The issue occurred when none of the fields in the document had shared_field_id values.
32.0.21 (30 Jun 2022)
Tasks
Updated
”View Task Queue” permission and filters in Task Queue – We’ve updated the View Task Queue permission to allow users to edit filters in the Task Queue tab.
Quality Assurance
Fixed
Transcription QA for Structured documents with checkboxes or signatures that have gone through Flexible Extraction – We’ve fixed an issue with completing Transcription QA tasks for machine-matched pages that had checkboxes or signatures that had previously gone through Flexible Extraction.
32.0.20 (24 Jun 2022)
User Experience
Updated
Custom login warning message – You can now create a custom login warning message, which is displayed when a user first logs in to your Hyperscience application. The message is shown as a dialog box that the user can dismiss by clicking the dialog's Continue button.
To add this message to your login experience, add the LOGIN_WARNING_BANNER_TEXT variable to your “.env” file, and enter your message as the variable’s value.
Tasks
Fixed
Viewing all active workers in the Active Workers card – We’ve fixed an issue that prevented users from viewing all active workers in the Active Workers card on the Tasks Overview page.
Transcription
Fixed
Rotating an image during Flexible Extraction – We’ve fixed an issue that resulted in an unexpected error when manually rotating an image during Flexible Extraction.
Submission Output
Fixed
Rejected documents in the Submission Output page – We’ve fixed an issue that caused documents that were rejected during Transcription Supervision to be accessible in the Submission Output page.
Models
Fixed
Displaying projected automation in the application – We recently changed how models exposed projected automation values to the application. These changes caused the following issues:
Table models' projected automation percentages were not displayed in the application in some circumstances.
The system crashed during the model-import process if the imported models used the former method of exposing these values.
This version contains fixes for both of these issues.
Recalibration of Classification models for Structured documents – We've fixed an issue that caused the page IDs of Flexible Extraction documents to point to pages without fields in some situations. This issue caused errors during the recalibration of Classification models for Structured documents.
Custom Supervision
Fixed
Decision choices and selecting transcription fields – We've fixed an issue that caused a keyer's decision choices to be cleared when they selected a transcription field.
Integrations
Fixed
Submissions created from the Box Folder Listener Block – If the folder specified in Folder to Move Completed Files did not exist, the Box Folder Listener Block created an infinite number of submissions. With the fix included in this version, the Box Folder Listener Block creates submissions only if the folder specified in Folder to Move Completed Files exists.
Authentication
Fixed
Logging SAML metadata-validation exceptions – We've updated how we log SAML metadata-validation exceptions to prevent the exception causes from being hidden.
Security
Fixed
Changing the javax.el library to jakarta.el and updating it – To address a security vulnerability, we’ve changed the Maven javax.el library to jakarta.el and updated it to 3.0.4.
32.0.19 (18 May 2022)
User Experience
Updated
Custom login warning message – You can now create a custom login warning message, which is displayed when a user first logs in to your Hyperscience application. The message is shown as a dialog box that the user can dismiss by clicking the dialog's Continue button.
To add this message to your login experience, add the LOGIN_WARNING_BANNER_TEXT variable to your “.env” file, and enter your message as the variable’s value.
Quality Assurance
Fixed
Transcription QA tasks illegible fields – We’ve fixed an issue that caused a field’s transcription value to be normalized to “None”. The issue occurred in Transcription QA tasks when the user reached consensus for a field that had been previously marked as illegible during Transcription Supervision or Transcription QA.
Health Check
Fixed
Cleaning up timed-out Health Check processes – We've fixed a rare issue that caused Health Check API requests to return limited and inaccurate information while the system cleaned up timed-out processes.
Security
Fixed
Updating jackson-databind and jackson-dataformat-cbor – To address security vulnerabilities, we’ve updated jackson-databind to 2.13.2.2 and jackson-dataformat-cbor to 2.13.2.
32.0.18 (11 May 2022)
Task Queue
Updated
Moving "Perform Tasks" in the user interface – We've removed the Perform Tasks button on the Task Queue page and added a Perform Tasks option to that page's Actions button. Clicking the Perform Tasks option has the same results as clicking the Perform Tasks button did in previous versions of Hyperscience.
Connections
Updated
"Timeout (Seconds)" setting for HTTP Notifier Blocks – We've added a Timeout (Seconds) setting to HTTP Notifier Blocks. This setting determines how many seconds the connection remains open if no data is received from the endpoint.
Definition of "Warm-Up Interval" – Previously, the Warm-Up Interval setting in the Universal Folder Listener was defined as the length of time that a file needed to remain unmodified before it became eligible for processing. With this update, Warm-Up Interval is defined as the length of time that must pass after a file's detection by the system before it becomes eligible for processing.
Reporting
Fixed
Keyer Performance Report end dates and daylight savings time – We've fixed an issue that caused the end dates of Keyer Performance Reports to be incorrect if the report's time period included the beginning or ending of daylight savings time.
Keyer Data Management
Fixed
Availability of page images – We've fixed an issue that prevented page images from loading in some situations.
32.0.17 (28 April 2022)
Model Validation Tasks
Updated
Order of Model Validation Tasks (MVTs) presented – We've changed the order of the MVTs presented to the user so that it matches the order in which they were generated by the system. Tasks generated first generally contribute more to the model's effectiveness, so finishing them first increases its effectiveness faster.
Submissions Table
Fixed
Adding the “Manual Classification” option to the “Status” filter – We’ve added the Manual Classification option back to the Submissions table’s Status filter.
Quality Assurance
Fixed
Use of system-level Transcription QA settings – We've fixed an issue that caused system-level Transcription QA settings to be used in place of flow-level settings in some situations.
Connections
Updated
"Audience" setting for HTTP Notifier and HTTP REST API Blocks – We've made the Audience setting optional in HTTP Notifier and HTTP REST API Blocks.
Authentication
Updated
Active Workers card and OpenID Connect usernames – If a user's first and last names are not retrieved by our OpenID Connect implementation, the user's email address is used in place of their full name in the Active Workers card on the Tasks Overview Page.
Security
Fixed
Updating com.google.code.gson to 2.8.9 – To fix a security issue, we’ve updated com.google.code.gson to 2.8.9.
Updating lxml to 4.7.1 – To fix an issue with Cross-site scripting (XSS), we’ve updated lxml to 4.7.1.
Updating OpenSSL and OS packages – When creating an installation bundle for a new version of Hyperscience, we now use the latest available version of OpenSSL, and we update relevant OS packages.
TVE Instances
Fixed
Database schema in the SQL Explorer tool – We’ve fixed an issue that prevented the database schema from being displayed in the SQL Explorer tool in TVE instances.
32.0.16 (8 Apr 2022)
Submission Processing
Updated
Rescheduling system tasks – The system no longer attempts to reschedule system tasks if there are parallel tasks that are timed out. This update reduces submission-processing times and increases system responsiveness.
Table Identification
Updated
Table Identification for large documents – Previously, you were always able to scroll through all pages of a document during Table Identification. As a result, Table Identification actions took between 2 and 5 seconds to load for documents with more than 18 pages. In v32.0.16, we’ve added a button (
) to the top toolbar that, when activated, restricts scrolling between pages and loads a single page at a time. Clicking this button improves the loading time for Table Identification actions for large documents. You can still navigate between pages via shortcut keys.
Data Types
Updated
Creating custom field data types (CFDTs) from patterns with “(space)” – Previously, if a user selected the (space) option in the Define Normalization dialog box when creating CFDTs from patterns, the “space” symbol was not visualized in the list of characters stripped in output. With the update included in this version, the “space” symbol is now visualized as “(space)” in the list of stripped characters.
Security
Fixed
Audit log editing permissions – We've fixed an issue that allowed System Admins to edit the audit log.
32.0.15 (24 Mar 2022)
Releases
Fixed
Release details page and problematic flows – We’ve fixed an issue with upgrading the application that caused release details pages to return error messages if releases were assigned to flows that did not have values for the layout_release_uuid field. These flows should not have been available for assignment. With the fix included in this version, upgrading the application no longer makes such flows visible in the Assign To Flow dialog boxes;, making them unavailable for assignment.
Assign to Flow button and Assign To Flow dialog box – We’ve fixed an issue that caused the Assign to Flow button to be enabled even if a release had been assigned to all available flows.
When a release was assigned to all available flows, the Current Release column in the Assign To Flow dialog box had an “N/A” value for each flow. A fix for this issue is included in v32.0.15.
Layouts
Fixed
Layout tags and layout names when importing existing layout variations – We’ve fixed an issue that prevented layout tags and layout names from being updated when importing existing layout variations.
32.0.14 (21 Mar 2022)
Models
Updated
Retrieval of model-type information – To improve system performance and prevent the creation of unnecessary Identification tasks, we've made model-type information more accessible to system processes.
Upgrades
Fixed
Migrating flows – We've fixed an issue that caused flow migrations to occur multiple times in a Hyperscience instance in some situations.
32.0.13 (16 Mar 2022)
Data Types
Fixed
Stripping of spaces in pattern-based data types – Previously, if (space) was selected as a character to be stripped from the normalized output of a pattern-based data type, spaces were not stripped from the normalized output. A fix for this issue is included in v32.0.13.
Models
Fixed
Model-import error messages in IE 11 – Previously, if an attempt to import a model in IE 11 was not successful, the error message shown was not completely contained in the dialog box. A fix for this issue is included in v32.0.13, and the user no longer needs to scroll the dialog box horizontally to view the complete message.
Model Validation Tasks
Fixed
Placement of bounding boxes – We've resolved a page-filtering issue in Model Validation tasks that caused bounding boxes to appear on incorrect pages.
Submission Processing
Updated
“Upload Submissions” dialog box and default layouts – The Upload Submissions dialog box no longer defaults to any particular Semi-structured layout.
Installations
Updated
PostgreSQL 12.10-alpine Docker image – Our installations now include PostgreSQL 12.10-alpine Docker images.
Upgrades
Fixed
Layout data in Oracle databases – We've fixed a SQL-optimization issue that caused a Too many bind variables supplied for this SQL statement error to occur when upgrading the application in environments with Oracle databases. This error occurred in instances where the database had a large amount of layout data.
32.0.11 (10 Mar 2022)
User Experience
Fixed
Font size of the “Archive layouts” dialog box’s text – We’ve made the Archive layouts dialog box’s text size consistent across the body of the dialog box.
Flows
Updated
Optimized queries for loading and calculating Transcription target accuracy and automation flow settings’ values – We’ve optimized the queries for loading and calculating the Transcription target accuracy and automation flow settings’ values. Previously, if a user edited a Classification setting’s value, the system would unnecessarily reload all target accuracy and automation values. With the improvement applied in v32.0.11, the system loads and calculates the Transcription target accuracy and automation values only in the following scenarios:
A user loads the Flow Studio page, and the Transcription Automation Training setting is enabled.
A user enables the Transcription Automation Training setting.
A user edits the Period of Records to Use setting’s value.
Flow Blocks
New
PDF Decrypt flow block – We’ve added a PDF Decrypt flow block that utilizes the QPDF command-line tool to decrypt PDF submissions prior to processing.
Models
Fixed
Optimized queries for loading the Model Details page – We’ve fixed an issue with executing unnecessary queries when opening the Model Details page for a particular layout. The issue resulted in system slowness.
Tasks
Updated
Counting and prioritizing overdue tasks – We've updated the calculations used to determine how many overdue tasks are in the system and the priority of each, reducing the loading time of pages in the Tasks section of the application.
Queries for Supervision Tasks, QA Tasks, and In Queue cards – We've merged and optimized the queries used to retrieve the data shown in the Supervision Tasks, QA Tasks, and In Queue cards on the Tasks Overview page. This update improves system responsiveness in instances with a large number of available tasks.
Submission Processing
Updated
Omitting unnecessary columns from queries – We've optimized the database queries run by the system during various user tasks. The queries retrieve data only from columns relevant to the task the user is performing, increasing system responsiveness in high-volume instances.
Documents
Updated
Showing data about documents on the Documents page – We've updated the queries used to show data about documents' submissions, task restrictions, and pending tasks on the Documents page of the application. These updates improve system performance in instances with a large number of documents.
Reporting
Fixed
Duplicate results for System Transcription Sampled Errors report – We’ve fixed an issue with the System Transcription Sampled Errors report that caused duplicate results to be displayed for each machine entry when a user filtered by System (Machine + Manual). With the fix applied in v32.0.11, the System Transcription Sampled Errors report displays only 1 result for each machine entry.
Optimized queries for loading the Transcription Sampled Errors report – We’ve fixed an issue with the Transcription Sampled Errors report that caused system slowness.
Time zones ahead of UTC time – Previously, when the system time zone is a time zone ahead of UTC time, some tooltips on the Reporting pages showed dates that were one day ahead of the data's actual dates. A fix for this issue is included in v32.0.11.
Output Blocks
Fixed
Configuration when an Output Block is the only block – We've fixed an issue that prevented the full configuration of Message Queue and HTTP Notifier Blocks if they were the only blocks in a flow (e.g., "Document Processing Notifications (V32)").
Databases
Fixed
Optimized database queries for scheduling recalibration, auto-thresholding, and finetuning – We’ve fixed an issue with executing unnecessary database queries for scheduling recalibration, auto-thresholding, and finetuning.
Submission Retrieval Store
Updated
Support for AWS Signature Version 2 in S3 submission retrieval stores – We've added support for AWS Signature Version 2 in requests sent to S3 submission retrieval stores.
SaaS
Updated
Autoscaling of trainers – We've updated the registration of trainers in SaaS instances to pass trainer-version data to the application, helping to optimize the autoscaling of trainers.
Security
Updated
TLS ".env" file variables and output connections – The values for the HS_TLS_CA_BUNDLE and HS_TLS_VERIFY_ENABLED ".env" file variables now apply to connections that use Java Message Service (JMS). These connections include RabbitMQ, ActiveMQ, and IBM MQ output connections.
For more information about these variables, see Security.
Fixed
Viewing HTML submissions – We’ve fixed an issue with viewing HTML submissions that allowed Cross-site scripting (XSS) in IE 11.
32.0.10 (17 Feb 2022)
Flows
Fixed
Optimized queries for opening flows - We’ve fixed an issue with executing unnecessary queries when opening a flow. The issue resulted in system slowness.
OpenID Connect
Updated
Parsing of OpenID Connect’s groups claims – We’ve improved how the system reads OpenID Connect’s groups claims by adding support for additional delimited strings. Previously, Hyperscience supported OpenID Connect’s groups claims only in the format of lists of strings.
32.0.9 (16 Feb 2022)
Submission Processing
Fixed
Compatibility of submission-processing tasks across versions – We’ve fixed a compatibility issue for some submission-processing tasks that caused:
trained ID models to become unused during submission processing, and all fields to go to Field ID Supervision, and
submission processing to fail when certain flow blocks were in use by a flow.
Training
Fixed
Field ID model training and large documents – We’ve fixed an issue that caused Field ID model training to fail when initiating training with large documents that have a large number of pages and long segments of text on each page.
32.0.8 (8 Feb 2022)
Submission Processing
Updated
Storing field data types in Submission objects – We’ve fixed an issue with executing unnecessary MSSQL queries for retrieving field data types during submission processing. These queries slowed down system performance. We now store field data types in Submission objects, which leads to increased performance.
Field Identification
Updated
Optimized payload queries for Field Identification QA tasks – We’ve removed redundant operations from payload queries for Field Identification QA tasks.
Infrastructure
Updated
Killing database sessions using local time – Previously, using local time in an Azure SQL Managed instance caused the system to measure blocker and sleeping database sessions as having lasted more time than they actually lasted. As a result, the system killed such database sessions as soon as they moved to a blocker or sleeping state. A fix for this issue is included in v32.0.8.
32.0.7 (7 Feb 2022)
Classification
Fixed
Classification tasks and system performance – Previously, the “Document Processing (V32)” flow generated unnecessary CPU-heavy tasks when the Semi-structured Classification flow setting was enabled and one of the following conditions applied:
The release contained only Structured layouts.
The Manual Classification Supervision flow setting was disabled, and there was no trained Classification model.
These CPU-heavy tasks caused processing delays. A fix for this issue is included in v32.0.7.
Databases
Updated
Moving transformed outputs to the file store – To reduce the amount of data in the database, we've moved transformed submission outputs to the file store.
Fixed
MSSQL queries and system performance – We’ve fixed an issue with executing unnecessary MSSQL queries that caused spikes in CPU usage. These spikes resulted in system slowness.
32.0.6 (27 Jan 2022)
We did not release v32.0.5. Its contents are included in v32.0.6.
Machine Classification Block
Updated
Structured document image transformations – To make Structured document image transformations available to Custom Code Blocks, we've added the transformation_matrix, template_image_shapes, and submission_image_shapes fields to the JSON output of Machine Classification Blocks.
Custom Supervision
Fixed
Order of page thumbnail images – We've fixed an issue that caused page thumbnail images to appear out of the order given in page_ids when group_by_documents was set to false.
Halting of submissions with pages matched to incorrect layout variations – Previously, when pages marked as matched to incorrect layout variations during Classification Supervision were sent to Custom Supervision, the pages' submissions would halt. A fix for this issue is included in v32.0.6, and keyers can view these pages when completing Custom Supervision tasks.
Halting of submissions without a documents key – We've resolved an issue that caused submissions to halt if they were sent to Custom Supervision and did not have the documents key in their Submission data objects.
32.0.4 (18 Jan 2022)
Flows
Fixed
Exporting flows without names – Previously, if a user exported a flow that did not have a name, a JSON named undefined.json was generated, but it was not downloaded to the user's machine. With the fix included in this version, users are able to export these flows, and the file names have the format of Flow - - .json.
Custom Supervision
Fixed
Inclusion of non-selected choices in decision selections – We've fixed an issue that sometimes caused non-selected choices in single-select decisions to be recorded as selected choices. For example, if the choices for a decision included "Valid" and "Not Valid," both choices would be recorded as selected.
Clearing of selected choices when clicking on transcription fields – We've resolved an issue that caused selected choices to be cleared when a keyer clicked on a decision that required transcription.
Task Queue field count for Custom Supervision tasks – We've fixed an issue that caused the value in the Fields column of the Task Queue to always be "0" in rows for Custom Supervision tasks.
Document-specific tasks in the Task Queue – We've resolved an issue that caused the system to create document-specific rows in the Task Queue for submissions containing Custom Supervision tasks. Because Custom Supervision tasks apply to submissions rather than documents, keyers could not take action on the tasks listed in these document-specific rows.
Training
Updated
Table Identification training enhancements – We've improved the Table Identification training processes to run in less time and generate fewer model-validation tasks (MVTs).
Databases
Updated
Domain accounts for SQL Server connections – You can now connect Hyperscience to SQL Server using domain account login credentials.
32.0.3 (14 Jan 2022)
Submission Processing
Fixed
Processing times and Transcription Supervision tasks for fields with multiple occurrences – We've fixed an issue that caused processing delays during Transcription Supervision for submissions with fields with multiple occurrences.
“View Cases” permission and uploading submissions – We’ve fixed an issue that prevented users who did not have the View Cases permission from uploading submissions in the application.
Supervision
New
Custom Supervision in embedded Supervision widgets – The embedded Supervision widget now supports Custom Supervision tasks.
Task Queue
Fixed
Custom Supervision tasks in the Task Queue – We’ve fixed an issue that caused a Custom Supervision task to appear multiple times in the Task Queue when a submission contained more than one document.
Permissions
Fixed
Submission-level task restrictions coming from the API – We’ve fixed an issue that prevented submission-level task restrictions that were set via the API from being applied to Supervision tasks.
Trainer
Updated
Changing the trainer’s mount directory – We’ve changed the trainer’s mount directory from media to trainer_media to prevent the trainer and the app from sharing the same mount directory.
Input Connections
Updated
“Headers to Include” setting for Email Listener – We’ve added a Headers to Include setting for the Email Listener input connector. This setting allows you to include headers from emails ingested into Hyperscience via the Email Listener connector.
Fixed
Hiding “Exchange” and “Routing Key” settings for RabbitMQ Listener – We’ve hidden the Exchange and Routing Key settings for RabbitMQ Listener. Previously, these settings were displayed for RabbitMQ Listener but should not have been.
Databases
Updated
Domain accounts for SQL Server connections – You can now connect Hyperscience to SQL Server using domain account login credentials.
32.0.2 (6 Jan 2022)
Languages
New
Support for Portuguese submissions – To support our expansion in Europe and Latin America, we've included a Portuguese language model in v32.0.2. Customers can now submit both Structured and Semi-structured documents in Portuguese, regardless of whether they contain printed or handwritten text.
Quality Assurance
Fixed
Inconsistent Table Identification QA task creation – Previously, Table Identification QA tasks would sometimes be created when they should not have been, and at other times, the system was prevented from creating these tasks as expected. A fix for this issue is included in v32.0.2, and the system creates Table Identification QA tasks consistently.
Reporting
Updated
Threading and scheduling of data-aggregation tasks – We've improved how the system distributes and schedules data-aggregation tasks for reports. With these changes, report-creation background processes can now be scheduled outside of business hours, and report generation is now faster and more reliable.
32.0.1 (16 Dec 2021)
Supervision
New
Custom Supervision – You can now tailor a Custom Supervision interface to a specific business process. This functionality allows you to build different and task-specific Custom Supervision interfaces for each flow.
Supported tasks in v32 include:
Validate transcriptions across multiple documents.
View external text.
Apply custom decisions to documents, pages, fields, and cases.
Manually assign documents to cases.
Note that you cannot customize QA interfaces.
Custom Supervision Block
To enable Custom Supervision for a specific flow, we’ve introduced a Custom Supervision Block. You can include a Custom Supervision Block in a custom flow with Hyperscience’s assistance. The only requirement is that a Custom Code Block should precede the Custom Supervision Block. For each custom flow, you can configure a different Custom Supervision Block with its own name and settings. In a Custom Supervision Block, you are able to define the following:
What data to show in a Custom Supervision task.
Which actions a user can complete in a Custom Supervision task.
Custom Supervision task’s page
We’ve introduced a three-panel page for Custom Supervision tasks.
Left panel - allows you to group page thumbnails by documents or cases. You can also choose to group the thumbnails by both documents and cases or to have no grouping at all.
Middle panel - shows a full page view of the document pages that are sent to Custom Supervision.
Right panel - allows you to configure up to three tabs that can supplement any of the supported tasks- validate transcription, view external text, apply custom decisions, and manually assign documents to cases.
You can access Custom Supervision tasks through the Tasks Overview page, the Task Queue page, the Submissions table, and the Cases table.
Note that data from Custom Supervision is not sent to QA and is not used in model training.
Data Types
New
New currency and email data types – We've added the following default data types to the system:
Currency data types for trailing signs – These data types allow keyers to express negative currency values by surrounding the value in parentheses or by entering a negative sign or "CR" after the value:
Currency Trailing Sign - X,XXX.XX
Currency Trailing Sign - X.XXX,XX
Separated Currency Trailing Sign - X,XXX XX
Separated Currency Trailing Sign - X.XXX XX
For example, with these data types, 200-, 200CR, and (200) are all equivalent to -200.
Email data type for international email addresses – The Email Address International data type uses a regular expression rather than a language model for validation. During normalization, all letters are made upper case, and no other changes are made. With this data type, email addresses are transcribed agnostic to the layout language selected, which leads to increased performance on documents that contain emails from multiple countries and languages.
Fixed
Existing ML Configuration drop-down list in IE 11 – We've resolved an issue that prevented the options in the Existing ML Configuration drop-down list from appearing when creating a data type in IE 11.
Creating custom field data types (CFDTs) from patterns – We’ve fixed an issue that prevented users from creating a pattern CFDT.
Creating a data type from a list of duplicate values – We've fixed an issue where users were allowed to create data types from a list of values containing duplicate values.
Layout Editor
Fixed
Unexpected warning when creating a Semi-structured layout – Previously, entering and deleting a layout name when creating a Semi-structured layout caused “A layout with that name already exists in your library.” to appear as a warning message. A fix for this issue is included in v32.0.1 and later.
Layouts
Updated
Enabling Find Potential Layout task creation with no trainer attached – The system can now create Find Potential Layout tasks even if there is no trainer attached. This functionality gives users the ability to shut down the trainer machine without affecting the creation and queuing of trainer tasks.
Fixed
“Save changes” without creating a layout tag – Previously, on the Layout Details page, entering a layout tag name and clicking the Save changes button without actually creating a tag resulted in a “Select_create” tag being added. A fix for this issue is included in v32.0.1 and later.
Archiving layouts – We've resolved an issue that prevented users from archiving layouts.
Models
Updated
Improved automation for tables – With this update, we've increased automation rates for documents containing tables. Specifically, we've improved our row-prediction capabilities, particularly in tables with 30 or more rows. As a result of these changes, we’re now able to predict the locations of rows that we were not able to predict accurately in previous versions.
Fixed
Table ID models and live releases that are assigned to multiple flows – Previously, if a live release was assigned to multiple flows, and the release contained a layout with a table that was also a part of other live releases, the Model Details page did not show information about the Table ID model, and the Run Training button was unavailable. A fix for this issue is included in v32.0.1 and later.
Ranges of potential automation – We've fixed an issue that prevented the range of potential automation rates from appearing on table models' Model Details pages.
Flows
Updated
Resizing of Flow Run diagrams – We've updated the resizing options for Flow Run diagrams to improve the user experience.
Fixed
Duplicate Folder Listener connections across flows – We’ve fixed a flow-validation issue that prevented the system from creating error messages when identical Folder Listener connections were added to different flows. Having Folder Listener connections with the same settings across flows causes resource-allocation issues and is not permitted.
Flow Studio
Fixed
Text in “Save and Deploy?” dialog box – We've edited the text in the “Save and Deploy?” dialog box to match the flow terminology used throughout the application, among other changes.
Submission Processing
Updated
Task polling for blocks – To reduce database queries, we've implemented a task-polling mechanism for blocks, which lets the system know the resources each block has available to complete tasks.
Fixed
Task synchronization across flow blocks – We've fixed a task-synchronization issue across flow blocks that caused submissions to fail, particularly in high-volume instances.
Clicking Back after completing a submission – We've fixed an issue that caused a 404 page to appear when a keyer clicked the Back button after completing a submission. Clicking this button now takes the keyer back to the Submissions page.
Table Identification
Updated
Excluding documents with many text segments from Table Identification QA – You can now prevent the system from generating Table Identification QA tasks for documents whose text-segment count exceeds the limit you specify. For assistance in setting this limit, contact your Hyperscience representative.
Fixed
Table Identification QA tasks and processing delays – We've fixed an issue that caused system resources to be consumed by Table Identification QA tasks. This issue prevented other tasks from being completed, resulting in submission-processing delays.
Deleting individual cells – We've resolved an issue that caused a cell's column to be deleted when a keyer right-clicked on a cell and clicked Delete Cell.
Opening and closing the Table ID keyboard shortcut sidebar in IE – We've fixed an issue that prevented users from opening the Table ID keyboard shortcut sidebar after opening and closing the sidebar once.
Distinguishing cells in identical locations across pages – We've fixed an issue that prevented the system from distinguishing cells in identical locations on different pages. This issue primarily occurred in low-quality images and with long segments of text.
Creation of Table Identification QA tasks after consensus – We’ve resolved an issue that caused the system to generate Table Identification QA tasks after consensus was reached.
Template-row identification
Fixed
Clearing a bounding box during template-row identification in Chrome and IE – We've fixed an issue where pressing the Esc key during template-row identification in Chrome and IE did not clear the selected bounding box.
“Next step” keyboard shortcut during template-row identification in Chrome and IE – We've resolved an issue where pressing the Enter or Return key during template-row identification did not complete template-row identification.
Find Missing Cells button appears during template-row identification – We've fixed an issue that caused the Find Missing cells for this column on all pages ( 
) button to appear during template-row identification, even though no rows could be automatically identified during that step of Table ID Supervision.
Shortcut for next step in Table Identification – We've updated the shortcut panel to show the correct shortcut for advancing to the next step of Table Identification (Control + Enter for Windows).
Table-identification review
Fixed
“Close task” keyboard shortcut during table-identification review in Chrome and IE – We've resolved an issue where pressing Option + Command + X for Mac or Alt + Control + X for Windows during table-identification review did not close the task.
Overlapping table cells from different rows – We've fixed an issue where if a table cell overlapped another cell from a different row, only one row was created instead of two.
Adding a table row to a multi-page document – We've resolved an issue where if you:
had a multi-page document with a table that spanned over two or more pages, and
you deleted all identified rows on a page that was not the first page, and
you clicked the + button to add a row,
dragging the bottom border caused that row to stay outside the table boundaries.
Table Identification QA
Updated
Excluding documents with many text segments from Table Identification QA – You can now prevent the system from generating Table Identification QA tasks for documents whose text-segment count exceeds the limit you specify. For assistance in setting this limit, contact your Hyperscience representative.
Fixed
Creation of Table Identification QA tasks after consensus – We’ve resolved an issue that caused the system to generate Table Identification QA tasks after consensus was reached.
Transcription
Updated
Entering Flexible Extraction transcriptions containing more than 2000 characters – In v32.0.0 and later, you can remove the 2000-character limit for text field transcriptions. With the update included in v32.0.1 and later, keyers can enter more than 2000 characters for a field during Flexible Extraction tasks.
Note that entering longer transcriptions may affect application responsiveness. To learn more, and to remove the character limit in your instance, contact your Hyperscience representative.
Fixed
“Next empty cell” keyboard shortcut and a hidden column during Table Transcription – We've resolved an issue where if the right-hand sidebar hid a table column during Table Transcription and a keyer pressed Control + E, the keyer wasn't moved to the next empty cell.
Cases
Updated
PII deletion's effect on Date last updated – When the system deletes PII from a case's fields, the Date last updated shown on the Case Details page is now updated to reflect when the deletion took place.
Training
Updated
Processing and scheduling of Field Identification QA and Model Validation Tasks (MVTs) – We’ve reduced the number of database queries related to Field Identification QA and MVTs, improving system efficiency in high-volume instances.
Reporting
Fixed
Legacy data in System Transcription Sample Errors report – Previously, when Legacy data was selected in the flows filter in the System Transcription Sample Errors report (Reporting > Accuracy), the report's data couldn't be downloaded. This issue has been fixed in v32.0.1 and later.
Keyer Data Management
Updated
Ordering of fields – We've made the ordering of fields on the Annotations page consistent across documents that share the same layout.
Time needed to save changes – We've reduced the amount of time it takes to save changes on the Annotations page.
Focusing on selected fields – When a field is selected on the Annotations page, the interface now brings that field into focus. This behavior matches that of the Supervision task interface.
Fixed
Pressing the Delete shortcut for a field without a bounding box – Previously, when a field without a bounding box was selected and the Delete shortcut was pressed, the Save button became enabled. However, there were no changes to be saved. A fix for this issue is included in v32.0.1 and later.
Save Changes button when moving to another document – Previously, when moving to the next document in the list, the user was notified that any changes they made would be lost if they did not save them before proceeding. However, when the user moved to the next document, any unsaved changes they made remained intact, and the Save Changes button was still enabled. This issue has been resolved in v32.0.1 and later.
Audit logs for annotation changes – We've reduced the size of entries in the audit log that are created from changes made on Annotation pages. This update prevents errors from occurring when annotation changes are saved.
Making changes when changes are being saved – We've fixed an issue that allowed users to make additional changes while previous changes were being saved.
Logs
Updated
Changing some error logs to info logs – We've moved some error logs to the info level because users could not take action on them, nor did they represent failure scenarios.
Image Deletion
Fixed
Adding image-deletion tasks to list of permitted tasks – We've resolved an issue that prevented image-deletion tasks from being permitted by the system in some cases. This issue caused unnecessary data to accumulate in the file store.
Input Connections and Output Connections
Updated
Logging details in UiPath error responses – We now log error messages from UiPath error responses, which contain details that can be helpful in debugging. Previously, we logged only the status codes in these messages.
API payload for UiPath Notifier Output connections – We've updated the API payload for UiPath Notifier Output connections to be consistent with the connections’ v28 payloads and the payloads of other output notifiers.
Fixed
Default names of connections – We've fixed an issue where if an imported connection did not have a name, the name defaulted to the internal name for the connection and the timestamp for its creation date and time, with all text strings connected by underscores. The name now defaults to the name of the connection type and the date the connection was created.
Universal Folder Listener's performance with a large number of resources – We've fixed a resource-detection issue in the Universal Folder Listener's block-process manager that caused delays when a large number of images were present in the source folder.
Validation of optional settings for Message Queue (MQ) connections – We've fixed an issue that caused validation errors when optional values were missing in the settings for MQ connections. This issue occurred when two MQ Listener connections were present in a flow.
Treatment of Group ID for FIFO Queues as a mandatory field – We've fixed an issue that caused the system to require a value for the Group ID for FIFO Queues in Amazon SQS Notifier settings, even though the setting is optional.
SSL CipherSuites and IBM Message Queue (MQ) connections – We've fixed an issue that caused input and output connections to IBM MQ servers to fail when SSL CipherSuites were enabled.
Installations
Updated
PostgreSQL 12.9-alpine Docker image – Our installations now include PostgreSQL 12.9-alpine Docker images.
Because previous installations included earlier versions of PostgreSQL Docker images, upgrading TVE instances with v28 or earlier to v30 or later requires a database migration. This migration is not required in production instances.
Infrastructure
Updated
Cron jobs polling – To increase performance and reduce dependency on Service Broker, we’ve updated the cron jobs channel to use polling instead of database-provided notification mechanisms (e.g., LISTEN/NOTIFY).
Health check for database-provided notification mechanisms – We've added a check for database-provided notification mechanisms (e.g., LISTEN/NOTIFY) to our system health check. This check also sends an alert when Service Broker is disabled in instances with MSSQL databases.
Authentication
Fixed
Creating users with third-party authentication providers – Previously, if an error occurred when creating a user through a third-party authentication provider, that user would not be able to log in, even after the error was resolved. A fix for this issue is included in v32.0.1 and later.
Trainer
Fixed
Running the trainer on a non-English Ubuntu operating system – We’ve fixed an issue that prevented users from running the trainer on an Ubuntu operating system that is configured with a non-English language.
Databases
New
PgBouncer – We now include PgBouncer in our installation bundles, which allows you to minimize database usage by limiting the number of database connections the application can create. PgBouncer is supported for PostgreSQL databases only and is recommended for instances that have more than 100 CPU cores across all application machines.
Updated
Changes to database-notification health checks and polling – To improve scalability, we’ve removed the health check for the database-provided notification mechanism in production instances, and we're enabling polling for job queues and block-process manager channels. The database-notification health checks remain enabled in TVE instances.
Storage of text-segment data – We now group text-segment data by page, improving scalability and system performance.
Deleting records of completed tasks at designated points – To reduce the amount of data stored in the database, we now delete records of completed flow tasks after their results are saved.
Moving large JSON files and flow backups to the file store – To increase database efficiency, we now store the following data in the file store rather than the database:
large JSON files from completed tasks
large JSON files from short-lived tasks
backups of completed flows
data generated during the automatic classification of Structured documents
Executing system tasks in memory – System tasks are now executed in memory rather than in the database, helping to reduce overall database load.
File Storage
Updated
Support for AWS Signature Version 2 in S3 file stores – We've added support for AWS Signature Version 2 in requests sent to S3 file stores.
SaaS
Updated
CSRF token in Client Libraries – We've added CSRF tokens to the Java, Python, and C# versions of our Client Library.
Metric for pending trainer tasks – In order for our SaaS deployments to scale trainers properly, we've added an internal metric that counts the number of pending tasks for each version of the trainer.
Fixed
Metrics for running and failed trainer tasks – We've added internal metrics that count running and failed trainer tasks for each version of the trainer, which prevents the task-management system from shutting down trainers that have tasks in progress.
API Account users listed on the Users page – We've fixed an issue that caused API Account users to be listed on the Users page in SaaS deployments of the application.
API
Updated
Support for Base64-encoded JSON data in submission creation – You can now send submission data in JSON format when creating submissions via the Submission Creation endpoint. To do so, include the Content-Type: application/json header in your request. When sending requests with this header, note that the request body has a different format than requests sent as multipart/form-data or application/x-www-form-urlencode.
32.0.0 (2 Dec 2021)
Flows
New
Flow-specific settings for Transcription Automation Training – In our efforts to support multiple lines of business in Hyperscience, we have added support for flow-specific target accuracies for transcriptions. This feature allows you to set accuracies that match the individual SLAs your organization has assigned for each flow, without sacrificing automation or accuracy in the other flows in your instance.
As part of this update, we have moved the Transcription Automation Training settings from the settings page (Administration > Settings) to the settings for each individual flow. The Projection Automation Based on Target Accuracy graph, however, remains on the Settings page. You can filter the graph's data to show the projected automation for a given flow.
We have also added the Finetuning Only For Trained Layouts flow setting, which allows you to restrict transcription automation training to trained layouts.
Because each flow has its own accuracy settings, each flow now also has its own transcription, or finetuning, model. However, models can also be shared across flows.
Debugging flows – You can now see the debugging data for each block a submission was processed through. To view this data, go to the Submission Details page, click Actions, and then click View Flow Execution.
In this debugging view of the submission's flow, you can access processing metadata for the submission, along with:
the input and output JSON for the submission,
the log of processing results for each block, and
JSON Input, JSON output, and logs for each block in the flow.
The View Flow Execution data helps you to not only debug halted submissions, but also to test flows in your development environments.
Updated
Dependencies and warning messages – When you edit a setting for a flow or block, the system now notifies you about any other settings that will be affected if you chose to save your changes. These notifications appear in the UI, and any affected settings will be in the same settings panel as the setting you are editing (i.e., in the same block, or also in the flow settings).
Flow Blocks
Updated
Improvements to connections – We've made the following enhancements to Input Blocks and Output Blocks, as well as connection settings in other blocks:
Proxy support for Amazon SQS – We've added support for proxy Amazon SQS connections.
OAuth2 error handling – We've improved OAuth2 error handling to make messaging clearer.
Folder Listener health check – We've fixed an issue that caused the Folder Listener health check to show a "Healthy" status even when the system couldn't access the folder.
Message Queue (MQ) JSON – We've resolved an issue that caused invalid JSON to be accepted and created through Message Queue connections.
Dynamic UI for connection settings – We've enhanced the connection settings UI to hide settings that are irrelevant to currently selected options. We've applied this dynamic UI to Email Listener, HTTP Notifier, HTTP REST API, Database, MQ Listener, and MQ Notifier settings.
Default ports for IBM MQ and RabbitMQ connections – We've resolved an issue that caused an error to occur when the default port numbers for IBM MQ and RabbitMQ connections were entered in the connection settings.
Username and passwords for IBM MQ connections – We've made the Username and Password fields optional for IBM MQ connections.
Input Blocks
New
Universal Folder Listener Block – We have completely rebuilt our existing Folder Listener input connection as the Universal Folder Listener Block.
This new block improves upon our current Folder Listener in the following ways:
Instead of supporting only one file per submission, submissions can now contain multiple files.
The root folder will always be included in file retrieval, preventing confusion during the configuration process.
The block supports the inclusion of case data and external IDs in metadata files.
We've made the UI for selecting file extensions and browsing folder directories more intuitive, leaving less room for error.
If you are currently using the Folder Listener input connection, your connection will not be automatically migrated to the Universal Folder Listener Block upon upgrading to v32. To use the block, you will need to add the block to your flow and configure its settings.
Layouts
Updated
Checkboxes and Signatures in Semi-structured layouts (Semi-structured only) – You can now add checkbox and signature fields to Semi-structured layouts. To support these new Semi-structured field types, the system creates field locator models for checkboxes and signatures and includes the fields in the generation of Field ID Supervision tasks. Keyers identifying these fields can use the same one-click bounding box functionality as they do for other fields.
In addition to Field ID Supervision tasks, the system also generates Field ID QA tasks for both checkboxes and signatures, as well as Transcription QA tasks for checkboxes.
Note that only binary checkboxes (i.e., checkboxes that can only be "checked" or "unchecked") are supported in Semi-structured documents, which matches our checkbox support for Structured documents. Semi-structured checkboxes now also have flow-level target accuracy, threshold, and minimum legibility settings, just as Structured checkboxes do.
To help you monitor the accuracy of checkbox and signature identification results, we've added checkbox and signature data to the Field Output Accuracy report (Reporting > Overview).
Releases
New
Multiple live releases – You can now have multiple live releases in your instance, allowing you to have a dedicated release for each of your lines of business. To ensure that submissions are processed against the correct releases, you can assign releases to specific flows.
Note that each flow can only have one release assigned to it, but each release can be assigned to multiple flows.
Submission Processing
New
Captured Image Enhancement (Semi-structured only) – Hyperscience now supports the processing of Semi-structured submission images captured with mobile devices, tablets, DSLR cameras, and other traditional document sources. During submission pre-processing, the system will correct the image-orientation issues typically found in these images before attempting to locate data.
We are releasing this feature in beta in v32, and it is disabled by default. You can enable it by selecting Enabled under Captured Image Enhancement in the Machine Classification Block settings. When enabling this feature, ensure that the Image Correction option in that block is also enabled.
Captured Image Enhancement is not supported for Structured or Additional documents. Furthermore, we do not recommend using it to process Semi-structured documents with tables.
The system will not create Supervision tasks for documents processed with Captured Image Enhancement. As such, using this feature may impact thresholding, model validation, and training.
Languages
New
Support for printed Structured submissions in Arabic – We now support automation on printed Structured documents written in Arabic. The Arabic language model allows our system to extract printed data from Arabic documents, and keyers can complete Transcription Supervision and Transcription QA tasks by entering Arabic text.
Note that we do not support the extraction of Arabic text from Semi-structured documents at this time, nor do we support the extraction of handwritten Arabic text.
Updated
Address improvements for EMEA and Americas – We've made improvements to our language models to better process address formats in the following countries in EMEA:
Germany
Spain
South Africa
We've made similar address improvements for these countries in the Americas:
Mexico
Argentina
Colombia
Canada (postal codes)
User Experience
New
Completing Supervision tasks in third-party systems – You can now embed Hyperscience into your organization's data-processing platform, allowing keyers to complete Supervision and Quality Assurance tasks without logging in to the Hyperscience application. After integrating the Hyperscience widget in your platform, keyers authenticate into Hyperscience through their API token, and they can complete both Hyperscience and non-Hyperscience tasks in a single user interface. Keyers must log in to Hyperscience at least once to complete tasks through the widget.
In addition to completing individual Supervision and Quality Assurance tasks, users with Knowledge Worker permissions can use the embedded Hyperscience interface to complete all tasks for a specific submission or case.
The widget is being released in beta as part of v32 and currently is only supported for on-premise installations.
Support for Microsoft Edge – We now support the use of Hyperscience in Microsoft Edge. Users accessing the application through Edge will have the same user experience as they currently do when using our other supported browsers.
For a full list of supported browsers, see Infrastructure Requirements and TVE (POC) Infrastructure Requirements.
Supervision Tasks
New
Identifying multiple instances of a field (Semi-structured only) – You can now capture multiple occurrences of a single field during Field Identification, Flexible Extraction, and Transcription Supervision. This identification of multiple values prevents data loss and reduces the need for exception handling downstream.
For example, if a patent application has an "Inventor" field, the keyer completing Field ID Supervision can identify one inventor listed in the document, indicate that there are other inventors that need to be identified, and then identify the next inventor in the document. A similar process applies when transcribing a document's data during Flexible Extraction Supervision. If multiple occurrences of a field are identified during Field ID Supervision, the system generates a separate Transcription Supervision task for each occurrence.
If a document contains multiple occurrences of a field, its Document Output page shows all occurrences of the field, and they are also included in the JSON outputs for the document and its submission.
Document Classification
Updated
Grouping of manually matched Structured pages – In v31, we introduced the ability to manually match pages to Structured layouts during Document Classification Supervision. After Document Classification, each manually matched Structured page became its own document. In v32, keyers can now choose to group a manually matched page in the same document as pages that the machine matched to the same layout.
Structured documents containing both manually matched and machine-matched pages are sent to Flexible Extraction Supervision for data transcription. During Flexible Extraction, keyers can view the shared field list for the layout's variations and transcribe any fields that are present in the document.
If a document contains both manually matched and machine-matched pages, its Document Output and Submission Output pages indicate how each page was matched.
Table Identification
Updated
Template tool improvements (Semi-structured only) – We've made the following enhancements to the Template tool:
Keyers can complete Identification tasks for tables that span many pages in a document, without delays or timeouts.
Keyers can re-run row predictions for specific pages, increasing efficiency.
We've improved the UI for editing cells and associating them with rows.
Keyers can edit the boundaries of multiple rows, columns, or cells at the same time.
We’ve updated the UI for splitting rows to make the process more intuitive.
New and improved table model (Semi-structured only) – The table model has been rebuilt with a new architecture that solves many of the issues seen in v31. Some of the notable changes include:
Ability to capture only parts of long lines of text as table cells – If there are some values that need to be extracted that are part of a larger line of text, the model can now output the exact words that need to be extracted, without the extra words around them.
Improved automation – Table model automation has increased across the board, with non-gridded table use cases seeing the most improvements.
Better generalization to unseen document types – The model can better automate the processing of document types that are not part of the training set, and errors in them that are not sent to Supervision are occurring less frequently.
Field Identification
New
Keyer Data Management (Semi-structured only) – In previous versions of Hyperscience, ground-truth data for field locator models was not accessible in the application, and models that could negatively impact automation rates were automatically deployed without notice. To give you more control over your field identification data and automation, we're introducing the following changes in v32:
Continuous Model Improvement disabled – The Continuous Model Improvement setting is now disabled by default. We recommend enabling it only if instructed to do so by your Hyperscience representative.
Field Automation Rates – You can now view automation rates for individual fields on the Model Details page, making potential ground-truth issues visible at a glance.
Training Data permissions – We've added two new permissions: View Training Data and Edit Training Data. These permissions are enabled by default for the System Admin and Business Admin permission groups.
List of training documents – Users with the View Training Data permission can view the new Training Data tab on the Model Details page. Clicking this tab allows the user to view a list of the model's training documents. Clicking on a document in the list shows the annotations made for that document.
Editing annotations – Users with the Edit Training Data permission can edit a document's annotations, which changes the ground-truth data of the model.
Training Status – Users with the Edit Training Data permission can also indicate whether the document should always or never be included in future model trainings, overriding any PII Deletion settings the document may be subject to.
Updated
Training data for Field ID locator models – Beginning in v32, Field ID locator models are only trained with data from model validation tasks (MVTs) and QA tasks by default.
You can also use data from Supervision tasks to train these models. To enable this option in your instance, contact your Hyperscience representative.
Table Transcription
New
Table Transcription Quality Assurance (Semi-structured only) – The system now generates Quality Assurance tasks for table transcriptions, which gives your organization the ability to monitor the quality of transcriptions provided by your keyers or the machine.
You can enable Table Transcription Quality Assurance in the flow settings and set a sample rate in the Table Transcription QA Sample Rate field. Transcriptions are sampled on the cell level, so if a document's cell is selected for QA, the system generates a Table Transcription QA task for that document. The keyer experience for these tasks is similar to that of Field Transcription QA tasks.
We've also updated our reports to include Table Transcription Quality Assurance data. More details can be found in the Reporting section of these release notes.
Reporting
Updated
Table Transcription Quality Assurance data (Semi-structured only) – We've updated the following reports to include Table Transcription Quality Assurance data:
Automation (Reporting > Overview), data visible in the Transcription filter
Field Output Accuracy (Reporting > Overview), data visible in the Table Transcriptions filter
Manual Accuracy vs. Machine Accuracy (Reporting > Accuracy), data visible in the Transcription filter
Supervision Volume (Reporting > User Performance), data visible in the Transcription filter
Checkbox and signature accuracy data (Semi-structured only) – We've added checkbox and signature data to the Field Output Accuracy report (Reporting > Overview). To view it, select the Semi-structured Documents filter, and then select either Checkbox or Signature from the field types filter.
Integrations
New
Box integration – If you are a customer of Box, you can use our new Box Folder Listener and Output Blocks to connect your Box account to Hyperscience.
With our dedicated Box blocks, you can retrieve data from your system of record and send extracted data back to Box as metadata, allowing you to query more data within the Box platform.
Infrastructure
New
Secrets Management – To enable the encryption of system-level secrets, we now support the integration of the following secrets-management solutions with Hyperscience:
CyberArk Conjur (versions 5.0 and later)
AWS Secrets Manager
HashiCorp Vault
When setting up your integration, you can choose to store any or all of the infrastructure credentials that are currently in your ".env" file in your secret-management solution. These credentials may include passwords and secret keys for your database, file store, and submission retrieval store.
Because SaaS instances don't have system-level secrets, these integrations are only supported for on-premise and private cloud deployments of Hyperscience.
To learn more about integrating a secrets-management solution into your instance, see Secrets Management.
Databases
Updated
Deprecating support for Oracle 12.1 – Beginning with v32 of the application, Hyperscience will no longer support the use of Oracle 12.1 databases.
Submission Retrieval Stores
New
Submission Initialization Block for AWS S3 – We've added a Submission Initialization Block, where you can enter your S3 submission retrieval store settings. This block contains the S3 Submission Retrieval Store field that appeared in the flow settings in SaaS deployments in v31+ and in on-premise / private cloud deployments in v31.0.1+. It also contains the S3 Submission Retrieval Endpoint URL, allowing you to view and modify this URL without opening the “.env” file for your instance.
Upon upgrading to v32, your S3 submission retrieval store settings will not be automatically migrated to this block, and none of your current flows will be affected by this change. The block does appear by default in the "Document Processing (V32)" flow, which is included in v32 but disabled when upgrading from a previous version.
Security
Updated
Stateful sessions – When users logged in to previous versions of Hyperscience, the system generated cookies to store information about user sessions. This stateless method of managing sessions resulted in security issues, particularly around session invalidation.
To improve system security, in v32, the system stores session information in the database. When a user logs out, the system deletes the information about the ended session. If that user is logged in to the system in another browser or device, those sessions remain active, which differs from system behavior in previous versions.
To learn more about authentication management for stateful sessions, see Application Authentication Overview.
API
New
Layout Endpoints – We've added Listing Layouts and Retrieving Layouts endpoints to our API, allowing you to retrieve information about layouts programmatically.
Responses from the Listing Layouts endpoint include metadata about each Layout in the list, along with its layout variations. You can also choose to include archived Layouts in the response.
The Retrieving Layouts data model includes the shared field list for the Layout, a list of the Layout's variations, and tables included in the Layout, among other data points.
See our API documentation for more details about Listing Layouts and Retrieving Layouts.
Updated
Deprecating API v3 – With the release of Hyperscience v32, we are deprecating v3 of our API. We will no longer add features or fixes to API v3, and we will sunset it in Hyperscience v33. At that point, it will be removed from our application.
If you are using this version of our API, we encourage you to use API v5, the latest version of our API.
To learn more about API v5, see our API documentation.
Known Issues and Limitations
See Known Issues and Limitations in V32 for a list of known areas for improvement in v32.0.0. We expect to resolve the issues in future versions of Hyperscience.