Hi Abilash Mathew, I haven't evaluated a per-character (or per-word) accuracy, but things look fair (80-90%? per char) from spot-checks. Obviously the accuracy is dependent upon the quality of the input, and sometimes you get a lousy 300dpi scan made from 200dpi printed fax ...
I do run pdf first through imagemagik, but that is more to get pagination than anything else as sometimes an entire patient history is in one giant pdf. I am not an expert on ocr or any particular tool or method. There is a lot of Tesseract discussion on the web if you can find it. Sean -----Original Message----- From: [email protected] [mailto:[email protected]] Sent: Thursday, October 19, 2017 1:20 AM To: [email protected] Subject: RE: OCR engine used [EXTERNAL] Sean, What is the accuracy that you get from OCR? We are at 60-70% accuracy. Most of the documents are 200 DPI ones. Also, are you using any other software like Matlab for the OCR pre or post processing. Thanks, Abilash Mathew -----Original Message----- From: Mathew, Abilash (Cognizant) Sent: Monday, October 16, 2017 8:37 PM To: [email protected] Subject: RE: OCR engine used [EXTERNAL] Thanks Sean fir the quick reply and providing the valuable information. Regards, Abilash Mathew -----Original Message----- From: Finan, Sean [mailto:[email protected]] Sent: Monday, October 16, 2017 8:17 PM To: [email protected] Subject: RE: OCR engine used [EXTERNAL] Hi Abilash Mathew, I have only used Tesseract. Unfortunately, no ocr is perfect. I am by no means an expert on Tesseract, but perhaps I can help to get you started ... There are tricks that you can use to get it to work better with medical notes (besides training on fonts). Possibly the most effective is using a whitelist of desired characters using tessedit_char_whitelist and a series of characters that doesn't include things like hash, dollar, bar ... Another is to add a wordlist that contains words pertinent to your domain. See: https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_tesseract-2Docr_tesseract_wiki_ImproveQuality-23dictionaries-2Dword-2Dlists-2Dand-2Dpatterns&d=DwIFAg&c=qS4goWBT7poplM69zy_3xhKwEW14JZMSdioCoppxeFU&r=fs67GvlGZstTpyIisCYNYmQCP6r0bcpKGd4f7d4gTao&m=qAmnDC76ixUPUO1C4SCKEG2rudGkfy63Rxe4RXQ8vH8&s=mdvTV4CsdGjAgIX6yNzNYCrkBuDVrNvOgxKiv-R9vxI&e= https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_tesseract-2Docr_tesseract_blob_master_doc_tesseract.1.asc-23config-2Dfiles-2Dand-2Daugmenting-2Dwith-2Duser-2Ddata&d=DwIFAg&c=qS4goWBT7poplM69zy_3xhKwEW14JZMSdioCoppxeFU&r=fs67GvlGZstTpyIisCYNYmQCP6r0bcpKGd4f7d4gTao&m=qAmnDC76ixUPUO1C4SCKEG2rudGkfy63Rxe4RXQ8vH8&s=w6BHFOtmh6VsGVBFaH2yhVLqxyezeW8ozgRhM67ImS0&e= https://urldefense.proofpoint.com/v2/url?u=https-3A__stackoverflow.com_questions_9568165_custom-2Ddictionary-2Dfor-2Dtesseract&d=DwIFAg&c=qS4goWBT7poplM69zy_3xhKwEW14JZMSdioCoppxeFU&r=fs67GvlGZstTpyIisCYNYmQCP6r0bcpKGd4f7d4gTao&m=qAmnDC76ixUPUO1C4SCKEG2rudGkfy63Rxe4RXQ8vH8&s=vqJa6rcFsmUgCotpp3fbfF6epW4WiHCJWugr4eFIyWs&e= https://urldefense.proofpoint.com/v2/url?u=https-3A__www.mail-2Darchive.com_tesseract-2Docr-40googlegroups.com_msg10100.html&d=DwIFAg&c=qS4goWBT7poplM69zy_3xhKwEW14JZMSdioCoppxeFU&r=fs67GvlGZstTpyIisCYNYmQCP6r0bcpKGd4f7d4gTao&m=qAmnDC76ixUPUO1C4SCKEG2rudGkfy63Rxe4RXQ8vH8&s=ORz4k4McDLmQa64dLEgFCE-oVBIW0LNNh2mVMb2T2Xk&e= Good luck, Sean -----Original Message----- From: [email protected] [mailto:[email protected]] Sent: Monday, October 16, 2017 10:13 AM To: [email protected] Subject: OCR engine used [EXTERNAL] Hi All, Can you guys give some of the OCR engines used for Medical record text extraction from images? I am currently using tesseract and seeing some text extraction quality issues. Thanks, Abilash Mathew This e-mail and any files transmitted with it are for the sole use of the intended recipient(s) and may contain confidential and privileged information. If you are not the intended recipient(s), please reply to the sender and destroy all copies of the original message. Any unauthorized review, use, disclosure, dissemination, forwarding, printing or copying of this email, and/or any action taken in reliance on the contents of this e-mail is strictly prohibited and may be unlawful. Where permitted by applicable law, this e-mail and other e-mail communications sent to and from Cognizant e-mail addresses may be monitored. This e-mail and any files transmitted with it are for the sole use of the intended recipient(s) and may contain confidential and privileged information. If you are not the intended recipient(s), please reply to the sender and destroy all copies of the original message. Any unauthorized review, use, disclosure, dissemination, forwarding, printing or copying of this email, and/or any action taken in reliance on the contents of this e-mail is strictly prohibited and may be unlawful. Where permitted by applicable law, this e-mail and other e-mail communications sent to and from Cognizant e-mail addresses may be monitored.
