Spaces:
Sleeping
Sleeping
| ## Annotation protocol | |
| ### 1. Know the task | |
| - The **XBRL reference formula** comes from this company's filed calculation | |
| linkbase and is treated as the formula ground truth. | |
| - Do not invent, validate, or rewrite the accounting formula. | |
| - Your task is only to check whether each XBRL parent and child is mapped to the | |
| correct row in the displayed PDF table. | |
| - Use only the displayed image, the XBRL-to-PDF reference mapping, and the | |
| extracted statement rows. Do not retrieve the filing, hidden provenance, or | |
| another annotator's answers. | |
| ### 2. Work independently | |
| - Use only your assigned slot: **Annotator 1** or **Annotator 2**. | |
| - Do not switch slots or inspect the other annotator's progress. | |
| - Judge every case independently. | |
| ### 3. Compare the row mappings | |
| Check the reference table term by term: | |
| 1. Does the XBRL parent label point to the correct PDF parent row? | |
| 2. Does every XBRL child label point to the correct PDF child row? | |
| 3. Does the extracted PDF label describe the same financial item as the XBRL | |
| label, allowing harmless wording differences such as note references? | |
| 4. Are the mapped rows in the displayed statement and the intended table | |
| context? | |
| 5. Are any rows missing, duplicated, cropped, or too ambiguous to identify? | |
| Keep these three sign concepts separate: | |
| - **XBRL operation:** the fixed calculation-linkbase weight, displayed as | |
| **Add (+1)** or **Subtract (-1)**. It must not be edited or re-annotated, | |
| and it is not a unit or display scale. | |
| - **XBRL fact sign:** the sign of the semantic fact stored in the filing. This | |
| sign is not necessarily visible in the rendered statement. | |
| - **PDF display sign:** parentheses or a printed minus sign used to present a | |
| deduction to readers. It is not a second calculation operation. | |
| For example, `Interest expense (233)` can store a positive XBRL fact of `233` | |
| and use **Subtract (-1)** in the calculation relation. Its contribution is | |
| `-233`, not `-(-233)`. Conversely, a cost may be printed as the positive number | |
| `9,855` and still use **Subtract (-1)**. Never apply the XBRL operation and the | |
| PDF display sign twice. Do not use visible arithmetic alone to reject a | |
| reference relation; displayed totals may also differ slightly because of | |
| rounding. | |
| ### 4. Use the paired image views | |
| - Start every case in **Masked**, which is also the main benchmark input. | |
| - Open **Original numbers** only when values help resolve a row, unit, duplicate label, or crop. | |
| - Switching views does not change the task: verify PDF-to-XBRL row mappings only. | |
| - Do not derive, validate, or rewrite the XBRL formula from the visible numbers. | |
| - A printed parenthesis or minus sign does not by itself make a mapping wrong. | |
| - The system records whether the original-number view was opened before saving. | |
| ### 5. Choose one alignment decision | |
| **Aligned** | |
| Every XBRL parent and child is mapped to the correct visible PDF row. Leave the | |
| pre-filled row mappings unchanged and do not select any Issues. | |
| **Misaligned** | |
| At least one XBRL term points to the wrong PDF row, and the correct row is | |
| visible in **All statement rows**. Select every applicable Alignment issue and | |
| correct only the wrong PDF row IDs. | |
| Use **Misaligned** only for an incorrect PDF row mapping. If every mapped row is | |
| correct but the XBRL relation itself still appears questionable, do not change | |
| row IDs and do not choose **Misaligned**. | |
| **Not verifiable** | |
| The mapping cannot be checked from the displayed evidence because a required | |
| row is missing, cropped, unreadable, duplicated beyond resolution, in a | |
| different table or context, or the XBRL reference itself appears questionable. | |
| Select an Alignment issue and explain the problem in Notes. Do not rewrite the | |
| formula; cases flagged as a possible XBRL issue go to expert review. | |
| For a questionable formula or weight, choose **Not verifiable**, select | |
| **Possible XBRL reference issue**, and describe the concern in Notes. | |
| ### 6. Correct a misalignment | |
| - Use **Mapped parent PDF row** only if the parent mapping is wrong. | |
| - In **Child row mappings**, keep every XBRL child label unchanged and edit only | |
| the **Mapped PDF row ID** cell for an incorrectly mapped child. | |
| - Keep one row for every XBRL child. Do not add or delete formula terms. | |
| - The system preserves every XBRL coefficient automatically. | |
| - If the correct PDF row is not available, choose **Not verifiable** instead of | |
| guessing. | |
| ### 7. Handle units and display scale | |
| - A caption such as **in millions** or **in thousands** is normally a common | |
| display scale for the monetary rows in that table. | |
| - A common display scale does not change an XBRL coefficient. Never use | |
| `1,000` or `1,000,000` as a coefficient. | |
| - USD, shares, percentages, and per-share values are different units. A unit | |
| mismatch may indicate that the XBRL term was mapped to the wrong PDF row. | |
| - If the needed unit or scale is cropped and prevents row identification, | |
| choose **Not verifiable** and explain what is missing. | |
| ### 8. Assign confidence | |
| - **High:** every referenced row and label match is clear. | |
| - **Medium:** the decision is supported, but one visual or wording cue is weak. | |
| - **Low:** substantial ambiguity remains; explain it in Notes. | |
| Before saving, confirm that you checked the parent and every child mapping. The | |
| goal is PDF-XBRL row alignment, not manual accounting-formula annotation. | |