Extract data to table Robin 2.0

Extract data to table Version 8 (Python)

Action group: Robin OCR

Description

The action recognizes the text from the document, divides it into blocks and returns it in a tabular form.

Action icon

Settings of parameters

Property	Description	Type	Filling example	Mandatory field
Parameters
Path to file	The path to the file to extract the data from. Supported formats: jpg, jpeg, bmp, png, tif, pdf.	Robin.FilePath		Yes
Page number	The page number in the document from which to extract data. If the field is empty, data will be retrieved from all pages.	Robin.Numeric		No
Language	The suggested language of the text to extract.	Robin.String		No
Additional language	An additional language required for document recognition.	Robin.String		No
Algorithm	If Text is selected, the action only recognizes text data. If «Table», the action only recognizes tabular data. If «Text and Table», the action recognizes any data.	Robin.String		Yes
Distance between words	The maximum distance between words in the document’s text data. Used to divide text into columns in the resulting table. The default value is 20 pixels.	Robin.Numeric		No
Distance between lines	The maximum distance between lines in the document’s text data. Used to divide text into lines in the resulting table. The default value is 1 pixel.	Robin.Numeric		No
Folder path	The path to the folder where the image of the document page will be saved with the overlay blocks into which the action has divided the data. To save the file, you also need to fill in the «File name» field.	Robin.FolderPath	C:\doc\img	No
File name	Title of the document page image with overlaid blocks (no extension). If the action retrieves data from several pages, then a separate file will be created for each of them, to the name of which an index will be added. To save the file, you also need to fill in the «Path to folder» field.	Robin.String		No
Overwrite	If the value is «true» and an image file with the same name, index, and extension exists in the specified folder, the new file will overwrite it. If «false», the file will not be overwritten and the action will return an error.	Robin.Boolean	true	No
Parameters	Additional parameters that affect the result and quality of text recognition.	Robin.String		No
Sign	If the value is «true», then the word «Part» will be added before each recognized tabular or text part with the ordinal number of this part. If the value is «false», a blank line will be inserted before these parts.	Robin.Boolean	true	No
Trained model	A file with a trained Tesseract model in the format .traineddata.	Robin.FilePath		No
Results
Table	A table generated from data retrieved from a source document.	Robin.DataTable
Path to image with blocks	Collection of paths to image files of document pages with overlaid blocks.	Robin.Collection

Special conditions of use

The action recognizes the text of the document and splits it into blocks depending on the specified input parameters (line spacing/word spacing) extracts data from text documents and saves them as a table. The source document may not contain a table.

You can input documents:

only the text layer of a pdf-document
images only
both text layer and images.

The operation of the action is based on the algorithm of document text block extraction. Words and lines of the document are combined into blocks based on the maximum distance between words and maximum distance between lines. These parameters are set in the input parameters of the action.

Examples of such documents: cashier's checks; documents containing tabular data without delimiters; documents containing solid text.

The logic of the action, depending on the setting of the "Algorithm" parameter: if "Table" is selected, the action will return text from tables only, preserving its table markup. If "Text", it will return everything, but for text from tables, it will ignore its markup (i.e. it reads all text from the source document, and if tables were found, the text from them will be extracted not by table markup, but by distance). If "Text and Table", it will recognize and return text from tables and plain text as separate parts, preserving the markup for tables.

Note that if the trained model does not work or is poorly trained, you should replace the values of the "Language" dropdown list with the values of the "Additional language" list.