We are aware of the following issues and limitations in v31.0.0. We expect to address them in future versions of Hyperscience.
Models
[Fixed in v31.0.5] Compatibility of Field Locator models across v31 versions – In v31.0.0-v31.0.4, the application can only use Field Locator models trained in v31.0.0-v31.0.4. For example, if you are running v31.0.4 or v31.0.0 of the application, you can use Field Locator models trained in v31.0.3, but not models trained in v31.0.5 or later. With the update included in v31.0.5 and later, a Field Locator model trained in any version of v31 will be compatible with the application.
Case Collation
[Fixed in v32.0.0] "Add to an existing case" and application responsiveness – When manually uploading a submission, clicking the Add to an existing case drop-down list can cause the application to become temporarily unresponsive when there are many Case IDs in the system.
"Unmatched pages" tooltip in IE 11 – When hovering over the "Unmatched pages" ? icon on the Case Details page in IE 11, the tooltip does not appear.
Attempting to complete tasks for "locked" cases – If a keyer attempts to complete tasks for a case that another keyer has started working on, the system redirects the keyer to the Submissions page without explaining why. Other keyers are only able to complete these tasks once the original task times out and is returned to the available task pool.
Knowledge Workers viewing layout variation information – If a user in the Knowledge Worker user group clicks on a layout variation's info icon on the Case Details page, the page reloads without showing the Variation Information window.
Viewing documents removed from a case – If a document is removed from a case, the system removes it from the list that appears when clicking Switch document in the document viewer. However, the document is still visible from its original document viewer URL.
Link name on "Waiting for work" page – When completing document-level Supervision tasks for a case, the system shows a "Waiting for work" page with a Document Overview link. Clicking this link takes the keyer to the Case Details page, not the Document Details page.
[Fixed in v32.0.0] Adding files with identical names in simultaneous submissions – Adding case files with the same filename in two or more submissions ingested at the same time results in the file not being added to the case. If the case already has a file with that name, the system removes the file from the case and does not add it back.
Notification for successful deletion of a page or document – The notification that appears after successfully deleting a page or document from a case hides the ⬅ Back link at the top of the page.
[Fixed in v32.0.0] Transformed submission outputs and cases arrays – The cases array is always empty in transformed submission outputs, even if cases include the submission's content.
[Fixed in v32.0.1] PII deletion and "Date last updated" value – When a PII-deletion job deletes the content of a field in a case, the case's "Date last updated" value on the Case Details page remains unchanged.
PII deletion and document status indicators – When a PII-deletion job deletes a document in a case, the document's status is not updated in the document viewer. Specifically, the colored indicator next to the document's ID remains unchanged, as does the status shown when clicking More details.
PII deletion and statuses of fields and unmatched pages – When PII-deletion jobs delete fields or unmatched pages in a case, their statuses on the Case Details Page still show the statuses from before the PII deletion.
Redirect after timing out of "Waiting for work" page – When completing a case's Supervision tasks and the "Waiting for work" page times out, the system redirects the keyer to the Submissions page instead of the Case Details page.
Machine Collation Blocks directly following each other in a flow – If a flow contains two or more Machine Collation Blocks without Custom Code Blocks separating them, the flow's submissions will be halted.
Table ID Supervision
System responsiveness with large documents – When a document contains 100 pages or more, the system can become slow to respond.
Workaround: Divide the submission into multiple smaller submissions of 10 pages or less, and bind the output together via a Custom Code Block.
Identifying cells that are split across pages – If cells in a single row are split across a page break, keyers won't be able to identify them in Table ID Supervision. The cells on the subsequent page will be considered part of a new row, and the keyer won't be able to connect them to the previous row.
Deleting the top or bottom row in the table – Deleting the top or bottom row in a table during Table ID Supervision does not move the table boundary to the next closet row. Instead, the boundary remains where the deleted row was, making the neighboring row larger.
Skipping row duplication for single-row tables – It is not currently possible to skip row duplication for single-row tables. In these tables, the template row is the only row to be identified, and no row duplication is required.
Adjusting the size of multiple cells – When a keyer has selected multiple cells, they can adjust only their width, not their height or position.
Selecting multiple cells – Keyers can only select multiple cells if they are part of the same connected range of cells. It is not currently possible to select multiple cells that do not share boundaries in the table.
Adding table rows in IE 11 – When creating new table rows by right-clicking the + button, the menu disappears when bringing the cursor to the Insert Rows text field, leaving only the cursor on the page.
Flexible Extraction
Reclassifying documents – While a keyer completing Flexible Extraction can indicate that a document's layout variation is incorrect, they are not able to reclassify the document.
One-click bounding boxes – One-click bounding boxes are not available when completing Flexible Extraction tasks.
Supported fields and tables – Keyers can transcribe only those fields and tables that are already part of the document's layout. As part of this limitation, Flexible Extraction does not support the transcription of multiple tables or multiple occurrences of a field.
Supported levels of transcription – Flexible Extraction supports only document-level transcription; case-level transcription is not supported.
Identification Supervision and QA
Keyboard shortcuts in IE 11 – After opening and closing the keyboard shortcut sidebar once, subsequent attempts to open the sidebar fail.
Workaround: Reload the page to open the sidebar again.
Table Transcription
Masking error with "Date" data type – Entering a single number into a cell with a "Date" data type (MM/DD/YYYY) results in an error message, preventing the keyer from entering additional numbers in the cell.
Submission Processing
Manual Classification and inaccurate submission statuses – When a submission has a Manual Classification task available, the system shows its status as "Processing" rather than "Manual Classification."
[Fixed in v31.0.1] S3 file storage and page statuses – Currently, if an instance has an S3 file store configured with missing or incorrect credentials, submissions containing files in the file store are still processed to completion. However, the system cannot retrieve the files from the file store, and the submissions should be halted. Specifically, the processing ends with the S3 pages having an "In Progress" status instead of the correct "Failed" status.
Field Normalization
[Fixed in v31.0.1] Normalization passes when it should fail – In submissions that contain fields with custom data type patterns, normalization is successful when manually transcribed values break the pattern. However, when the pattern is broken during Manual Transcription, normalization should fail.
Field Exceptions
Reporting of "Required Field Missing" exception – The system does not report "Required Field Missing" exceptions at the submission level for submissions processed in v30. Therefore, filtering by this exception type on the Submissions page creates an incomplete list of submissions, and the field exception does not appear at the submission level in the submission's JSON output.
API
Identification accuracy metrics in Manual vs. Machine Accuracy Report – The Manual vs. Machine Accuracy Report, whether viewed in the application or retrieved via API, does not have separate metrics for Field ID accuracy and Table ID accuracy.