Accessing this feature
Your access to the feature described in this article depends on your license package and pricing plan.
To learn which features are available to your organization and how to add more, contact your Hyperscience representative.
Table Identification tasks are used for Semi-structured documents with tables. In these tasks, you confirm or correct predicted table cells. If predictions are not available, you identify a representative row and the system uses it to predict the remaining rows and cells.
A table contains rows, columns, and cells. Table Identification supports regular tables, multiple tables in the same document, and nested tables. For layout configuration, see Creating Semi-structured Layouts and What is a Nested Table?.
Before you begin
Table Identification is available only for Semi-structured layouts that contain table-column definitions. Before processing documents:
Create the table and its columns in a Semi-structured layout.
Add the layout to a release and deploy the release.
Upload documents through Submissions or {{glossary.Training Data Management (TDM)}}.
For model-training and TDM workflows, see TDM for Identification Models and Training an Identification Model.
Start a Table Identification task
Go to Tasks > Perform Tasks.
In Supervision Tasks, find Identification, and then click Perform Tasks.
If the Table Identification guidance appears, review it and continue. You can choose not to show the guidance again.
If a document contains both fields and tables, you complete Field Identification before Table Identification.
Table Identification interface
The task interface contains the document pages and table annotations in the main area and the table controls in the right panel. The right panel shows the current table, its columns, table progress when the document contains multiple tables, Document details, and the action for moving to the next step or completing the task.
Column labels can indicate New Column, Missing column, Misplaced column, or an anomaly count. Treat these indicators as review prompts, but check every cell even when no warning appears.
Identify a template row
If cell predictions are available, the task opens directly in the review step. If predictions are not available, identify one representative row first.
Select a row that contains values for as many columns as possible. The row does not need to be the first row in the table.
Select a column in the right panel, and then identify the corresponding cell in the row. Repeat for every column that is present.
Click Continue to Review. The system uses the template row to predict the remaining table rows and cells.
The predicted rows are a starting point. Review every row and cell before completing the task.
Review table cells
Select a column or cell and confirm that each bounding box contains the complete value. Dashed yellow borders indicate that a cell may be incorrect, but cells without a warning must also be reviewed.
If... | Then... |
|---|---|
The bounding box contains the complete cell value. | Continue to the next cell. |
The box is in the correct location but excludes part of the value. | Drag the box edges or corners until the entire value is included. |
Adjacent text segments belong to the same cell. | Draw a box around all segments that belong to the cell. |
The prediction is in the wrong location, or no box appears. | Hold the spacebar and drag to draw a bounding box around the correct value. |
To discard the generated rows and start again from a different template row, click Manually Re-identify Table.
Manage rows and cells
Split a table into rows
If one boundary contains multiple table rows, use the row-splitting tool:
Extend the table boundary to include the required rows.
Enable Split row.
Select each position where a row boundary should be created.
Add or remove rows
Use the add-row control to insert a row. You can also open a row’s menu and select one of the available actions:
Insert 1 above
Insert 1 below
Delete selected row or Delete selected rows
Delete all rows above
Delete all rows below
Work with a column
Column-label actions let you:
Find Missing Cells for this column on all pages to generate cells that may be missing from identified rows.
Select all column cells on page to adjust the current page’s cells together.
Delete all column cells on page to clear the current page’s cells for that column.
Work with multi-page tables
Review every page that contains part of the table. If a row continues onto the next page, click Extend row [row number] to next page between the pages. The continuation is shown on both pages.
For documents with many pages, Scroll freeze can improve performance, but it disables continuous scrolling. Use the page controls to move between pages while it is enabled.
Work with multiple and nested tables
When a document contains multiple tables, the progress indicator shows the current table and the total number of tables. Complete each table in turn. The completion button shows Continue to next table until you reach the final table.
For a nested table, identify the child table first and then the parent table. The task titles and progress indicator identify whether you are working on the child or parent table. Make sure each child row is associated with a parent row.
Resolve completion warnings
The task checks the table structure before submission.
Message | Required action |
|---|---|
Duplicate cells | Make sure each row contains no more than one cell for the same column. |
Empty Row | Add at least one cell to every row or delete the empty row. |
Orphan Rows | Associate each child-table row with a parent-table row. |
Review All Pages | Return to the document and review the remaining pages, or confirm that all pages have been reviewed. |
Handle an incorrect layout
If the document was matched to the wrong layout:
Expand Document details in the right panel.
Click Mark Layout Incorrect or Mark Layout Variation Incorrect, depending on the available reprocessing behavior.
Review the confirmation message and continue.
If the flow contains an enabled Reprocessing Block, the incorrect layout is removed from the document’s pages and a Manual Classification task is created. Otherwise, processing for the document ends and no data is extracted.
You can also use Reject Document when the document should not continue through processing. Rejecting a document dismisses its remaining Supervision tasks.
Complete the task
After reviewing the current table, click Continue to next table when more tables remain. On the final table, click Submit Document. Resolve any completion warning before submitting.
After Table Identification, the identified cells can proceed to Transcription.
Keyboard shortcuts
Press F2 to open the keyboard shortcuts panel. The panel shows only the shortcuts that apply to the current Table Identification step and operating system.
The updated shortcuts are optional in v43.2 and can be enabled or disabled from the keyboard shortcuts panel. Use the panel as the source of truth instead of relying on a static shortcut list.