Table recognition reconstructs rows, columns and cell contents from an image or document. It combines reading the text with detecting its structure so that the data can be used in a spreadsheet editor.
When a table output is useful
If a report or register exists only as a scan and you need its values for calculations, choose an available CSV or XLSX conversion. Ordinary text does not store cell boundaries. For a PDF containing several tables, decide which parts you will need to verify after conversion.
Why the structure may change
Recognition relies on text positions, lines and gaps between blocks. Faint borders, skewed pages, merged cells and long labels make boundaries harder to detect. Words from one cell may be separated, while neighbouring cells may be combined. Detecting table structure does not guarantee an error-free output.
Checking CSV or XLSX data
Compare row and column counts, headings, record order and whether each amount belongs to the correct label. Check minus signs, decimal separators, percentages, leading zeros and similar characters separately. CSV delimiter and encoding settings in your editor can affect its display, so compare the data rather than appearance alone.
What a scan cannot restore
An image shows final values rather than the original spreadsheet formulas. Recognition does not automatically recover hidden sheets, formulas or the original formatting rules. Correct detected errors and verify key values before calculating. A clearer source and the correct OCR language help, but do not replace comparison with the original.