Document to JSON Extractor (Arabic and English)
Pricing
from $3.00 / 1,000 page processeds
PDF, DOCX, image or plain text in, structured JSON out, with the page and paragraph anchor for every value so a person can verify it. Arabic and English: Hijri dates, Arabic-Indic digits, right to left order and Arabic clause numbering all handled. Pass a JSON Schema to get exactly your fields.
Mode
mode
Optional
Which mode produced this row. analyze returns the whole document as text, paragraphs and clauses. extract returns only the fields your JSON Schema asked for.
Source type
sourceType
Optional
How the document reached the Actor: a URL, an uploaded or base64 payload, inline text, or one of the bundled samples.
Source reference
sourceRef
Optional
The URL or sample name the document came from.
File name
fileName
Optional
Name of the document that was read.
Format
format
Optional
Detected input format of the document.
Text source
textSource
Optional
Where the text came from. pdf-text-layer means the PDF carried real text; ocr means the page had to be read optically, which costs an ocr-page event instead of a page event.
Bytes
byteSize
Optional
Size of the source document in bytes.
Pages
pageCount
Optional
Total number of pages in the document.
Text pages charged
pagesCharged
Optional
How many pages were billed as the page event, meaning they had a usable text layer.
OCR pages charged
ocrPagesCharged
Optional
How many pages were billed as the ocr-page event. OCR costs roughly seven times a text-layer page, so it is priced separately.
Paragraphs
paragraphCount
Optional
Number of paragraphs the document was split into.
Clauses
clauseCount
Optional
Number of numbered or headed clauses detected, including Arabic clause numbering.
Characters
characterCount
Optional
Total character count of the extracted text.
Script
language
Optional
Dominant script detected in the document.
Direction
direction
Optional
Reading direction of the document, left to right or right to left.
Arabic characters
arabicChars
Optional
Count of Arabic letter characters in the document.
Latin characters
latinChars
Optional
Count of Latin letter characters in the document.
Normalisations
normalisations
Optional
Which normalisations were applied, for example Arabic-Indic digits folded to ASCII or Hijri dates converted.
Warnings
warnings
Optional
Non-fatal problems found while reading the document, such as low OCR confidence.
Extracted at
extractedAt
Optional
ISO 8601 timestamp of when the document was processed.
OCR confidence
ocrMeanConfidence
Optional
Analyze mode, OCR only. Mean OCR confidence across the page, 0 to 100. Low values are worth a human check.
OCR languages
ocrLanguages
Optional
Analyze mode, OCR only. Which language models the OCR pass used.
Full text
text
Optional
Analyze mode. The whole document as plain text, in reading order.
Paragraphs
paragraphs
Optional
Analyze mode. Every paragraph with its anchor, page number, direction and text, so any value can be traced back to where it appeared.
Clauses
clauses
Optional
Analyze mode. Every clause with its anchor, numbering label, heading and text.
Extracted fields
fields
Optional
Extract mode. An object holding exactly the fields your JSON Schema requested, keyed by field name.
Field reports
fieldReports
Optional
Extract mode. Per requested field: whether it was found, the parsed value, the raw text it came from, and the page and paragraph anchor that is the evidence for it.
Fields requested
fieldsRequested
Optional
Extract mode. How many fields your JSON Schema asked for.
Fields found
fieldsFound
Optional
Extract mode. How many of the requested fields were located in the document.
Missing required
missingRequired
Optional
Extract mode. Names of required fields that could not be found.
Complete
complete
Optional
Extract mode. True when every required field was found.
