OCR, data entry and keying services
Scanned books, forms, records and handwritten pages turned into accurate, editable text and structured data, in any script. Software does the first pass, and people key and proofread what it gets wrong.
Modern printing began in Europe in the fifteenth century.
The type was set by hand, one letter at a time.
भारत एक विशाल देश है।
ModemModern printing began in Europe in the fifteenth century.
The type was set by hand, one letter at a tirnetime.
भारत एक विशलविशाल देश है।
Example: OCR misreads "rn" as "m" and drops a Hindi vowel sign. A person fixes both.
- You send
- Scans, photos or PDFs of books, forms, records or handwriting
- You get back
- Word, searchable PDF, Excel, CSV, XML or JSON
- Checked by
- Double keying or proofreading against the source
- Priced by
- Page or record count, script and source quality
Where OCR stops and people take over
OCR (optical character recognition) software reads printed text from images. It's fast, but it makes predictable mistakes: "rn" read as "m", 1 and l mixed up, accents and vowel signs dropped, and columns or tables read in the wrong order. Data entry is the human work of keying and checking the text and fields that software can't be trusted with.
We do both, in Latin, Cyrillic, Arabic, Devanagari, Chinese and other scripts, with people who read the language.
OCR and data entry work
For publishers, archives, researchers and operations teams.
OCR with human correction
Printed books, reports and archives run through OCR, then proofread line by line against the scan.
Double keying
Two people key the same text independently. Software compares the results, and a third person settles every difference.
Handwriting transcription
Letters, notebooks, field records and forms filled in by hand, transcribed by people who read the hand and the language.
Forms and records
Fields from invoices, applications, surveys and registers entered into a spreadsheet, database or your template.
Non-Latin scripts
Arabic, Hebrew, Devanagari, Bengali, Tamil, Chinese, Japanese, Korean and Cyrillic keyed by native readers.
Proofreading
Existing digital text checked against the original pages, with every correction logged.
Accuracy you can check
We agree an accuracy target with you and measure against it on every batch.
- Double keyingTwo independent passes, compared by software, with every difference resolved by a person.
- Field rulesDates, totals, IDs and postcodes checked against format and range rules as they're entered.
- Sample auditsA random sample of each batch re-checked against the source, with the error rate reported.
- Unreadable textMarked as [illegible] with its location, never guessed.
How a keying project runs
Start with sample pages, so the quote and the accuracy target are based on your real material.
Sample pages
Send a few typical pages or records, and tell us the output format and fields.
Quote and target
A fixed price per page or record, a delivery schedule and an agreed accuracy target.
Capture
OCR or keying, depending on the source, by people who read the script.
Check
Proofreading or double-key comparison, field rules, and a sample audit of every batch.
Second check QADelivery
Files in your format, with the audit result and a list of anything illegible.
Formats
- Sources
- PDF, TIFF, JPG and PNG scans, phone photos, or printed books
- Text output
- Word, plain text, searchable PDF, or EPUB through eBook conversion
- Data output
- Excel, CSV, XML, JSON or your import template
- Scripts
- Latin, Cyrillic, Greek, Arabic, Hebrew, Devanagari and other Indian scripts, Chinese, Japanese and Korean
Related services
Questions about OCR and data entry
Can you key text in non-Latin scripts?
Yes. Native readers key and proofread Arabic, Hebrew, Devanagari, Bengali, Tamil, Chinese, Japanese, Korean, Cyrillic and other scripts, where OCR software is least reliable.
How do you check accuracy?
Depending on the project, by proofreading against the source or by double keying, where two people key independently and every difference is resolved. A random sample of each batch is then audited and the error rate reported to you.
Can you transcribe handwriting?
Yes. Handwritten letters, notebooks, records and forms are transcribed by people who read the handwriting and the language. Words that can't be read are marked rather than guessed.
What is the difference between OCR and data entry?
OCR is software reading text from an image. Data entry is people keying text or fields by hand. OCR is quicker for clean print. Data entry is more accurate for handwriting, poor scans, complex forms and scripts OCR handles badly. Most projects use both.
What do you need from us to quote?
A few typical sample pages or records, the output format and fields you need, and the volume. We quote a fixed price per page or record with an agreed accuracy target.
Send us sample pages
Send a few typical pages or records with the output you need. We reply with a fixed price per page or record and an accuracy target.
- Phone
- +44 20 3290 0116
- Phone hours
- Weekdays 9:30 to 16:00, weekends 9:30 to 12:00, UK time