DIRECT ANSWER / 07
How are ambiguous and repeated amounts handled?
An amount whose punctuation could be read two ways stays pending until a person chooses. An amount that appears in several paragraphs is extracted once per occurrence, so a figure updated in two places and missed in a third is visible rather than averaged away.
Why this review exists
Punctuation and repetition are the two ways a numeric review goes quietly wrong: one by reading a number incorrectly, the other by reading only the first copy of it.
Ambiguous punctuation
A figure written as 1.500 means one and a half in one convention and fifteen hundred in another. The review setting records which decimal style the documents use, and the sample uses a dot decimal with comma delimiters, but a claim that remains genuinely ambiguous is marked as such and stays pending.
A pending ambiguous claim cannot be mapped and cannot contribute to a rollup, so it cannot silently change a total. It appears in its own list and in the handoff packet as an ambiguous claim.
Resolving it is a person reading the source document, not the tool applying a heuristic. Guessing here would be worse than leaving the claim open, because the guess would look like a confirmed reading.
Repeated amounts
Claims are extracted per occurrence rather than per distinct value, so a figure that appears in three paragraphs produces three claims with three positions and three contexts. That is exactly what a reviewer wants when a budget is revised: the amount was updated in the table and in two paragraphs, and the third still says the old number.
In the sample, 46,500 appears once in the prose as the restated proposal total and once as the requested amount on the review brief, and the reviewer marked the narrative occurrence as a rollup rather than a line-item claim.
Nothing is deduplicated. The tool shows the occurrences and their contexts and leaves it to a person to say which are the same fact and which are different ones.
Practical sequence
- Set the decimal style the documents actually use
- Review every claim marked ambiguous before mapping anything
- Read the repeated-amount view for figures that appear more than once
- Confirm that every occurrence of a changed figure was updated
What remains outside the app
Grantweft performs arithmetic and textual cross-reference on user-supplied material. It does not write proposals or determine eligibility, allowability, accounting treatment, compliance, completeness, accuracy, submission readiness or funding outcome.
The private desk keeps exact source labels, stable budget row identities, paragraph-based claims, confirmed mappings, visible arithmetic, findings, human decisions, review freezes and passive handoff files connected without an account or remote application database.