- document performance
- document type performance
Document performance
In an ideal performance scenario, you extract data from digitally generated PDFs using text-based or coordinate-based Sensible methods, such as Label, Row, Region, Text Table, and Document Range. In order of largest to smallest impact, these factors add seconds to the ideal document processing time: Over 10 seconds per documentFactor | Notes |
---|---|
whole-document OCR for document images | Sensible takes 10 seconds or more to OCR an entire document. You can speed OCR up for documents that are 25 pages and shorter by choosing Sensible’s Google OCR option for a document type. |
whole-document table recognition | Avoid configuring Sensible to search a whole document for tables. Instead, configure a table stop. For examples, see any of the table methods. |
Factor | Notes |
---|---|
OCR preprocessor | Some documents mix digital text with text images, for example by embedding scanned pages in a digital PDF. Speed this up by OCRing select pages, not the whole document. For more information, see the OCR preprocessor. |
Fixed Tables with stops,Nearest Checkbox method | Sensible processes Fixed Tables that include a Stop parameter in less than 5 seconds. Or, convert to the faster Text Table method, which skips table recognition. |
LLM-based methods | For methods based on large language models (LLMs), see Sensible Instruct |
Factor | Notes |
---|---|
Box method | To improve processing speed, convert the more flexible Box method to the strictly coordinate-based Region method. |
Signature method, Checkbox method,image coordinate extraction | These methods have no alternatives. See the following section for ways to avoid running these methods except when necessary. |