Skip to main content
Classification proposes where an uploaded document belongs in the deal’s taxonomy. Plomo analyzes the source, assigns a category and subcategory when it can, and records confidence and supporting evidence. Search indexing is tracked separately. A source may already be searchable while classification is pending or has failed.

​
Read a classification

Open a source in Sources to inspect its assigned category, available evidence, and processing details. The confidence score is used to prioritize review; it is not a measured probability that the result is correct for your particular document. The classification pipeline uses these review thresholds: These are classification states, separate from the review and clearance status of an information request. An accepted classification is not a human approval or a finding that the source is complete.

​
How classification runs

Text-based analysis uses extracted content and the taxonomy. PDF processing can also use rendered page images to interpret scanned or visually structured material. The processing path determines which inputs are used; every file does not necessarily receive both forms of analysis. The pipeline validates model output and can make a bounded number of recovery attempts for malformed or unusable responses. A low-temperature initial call reduces variability, but repeated runs are not guaranteed to produce identical results. Some files still need manual correction or a retry.

​
Correct an assignment

Inspect the source and the evidence first. If it belongs elsewhere, use Move to in the source’s menu and choose the correct category. Available actions depend on your editing permissions. After a correction, inspect related items in Coverage Map. Category placement and evidence coverage are different decisions; do not assume a category move establishes that every request in that category is satisfied. For the wider workflow, see reviewing documents.