Preprocessors

Use the following preprocessors to clean up your documents before extracting structured data. Preprocessors execute in the order you define them in an array.

PreprocessorImageNotes
DeskewClick to enlargeCorrects the alignment of documents that are skewed, for example as a result of being photographed at an angle instead of straight on.
LigatureClick to enlargeIntelligently replaces Unicode ligatures in a text extraction.
Merge LinesClick to enlargeCorrects oversplit lines.
NLPAdvanced prompt configuration for all the large-language model (LLM)-based methods in a config.
OCRClick to enlargeSelectively OCRs pages in documents containing a mix of digitally generated text and text images (such as scanned text). If the whole PDF is a scan, you don't need to configure this preprocessor.
Page FilterFilters out low-scoring pages given a bag of target terms and stop terms.
Page RangeIgnores pages outside the start page and end page.
Remove HeaderClick to enlargeRemoves repeating elements at the top of the page. Ignores header elements that overlap with the page's main body.
Remove FooterClick to enlargeRemoves repeating elements at the bottom of the page. Ignores footer elements that overlap with the page's main body.
Rotate pageClick to enlargeIn most cases, Sensible corrects page rotation automatically. If it doesn't, configure this preprocessor.
ScaleClick to enlargeCorrects the size of text in documents whose size varies, for example as a result of being scanned or photographed at different scales.
Split LinesClick to enlargeCorrects undersplit lines.