[
https://issues.apache.org/jira/browse/TIKA-3431?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17355788#comment-17355788
]
Sal edited comment on TIKA-3431 at 6/2/21, 3:08 PM:
----------------------------------------------------
I think I understand what's happening now. After testing some more, I found
that setting 'ocr_and_text_extraction' converts each PDF page to image and runs
OCR on it AND extracts text from the page. The output for each page is a
duplicate containing the OCR'ed portion followed by the text portion. This is
why those settings take so long to complete.
What I wanted was to extract the text and OCR only the inline images, not OCR
the entire page. So that requires the following settings
Content-Type: *application/pdf*
X-Tika-PDFextractInlineImages: *true*
X-Tika-PDFOcrStrategy: *auto* or *no_ocr*
I was under the impression that ocr strategy was something to do with how Tika
extracted text in general (i.e. applied OCR to the inline images or not)
was (Author: sallas):
I think I understand what's happening now. After testing some more, I found
that setting 'ocr_and_text_extraction' converts each PDF page to image and runs
OCR on it AND extracts text from the page. The output for each page is a
duplicate containing the OCR'ed portion followed by the text portion. This is
why those settings take so long to complete.
What I wanted was to extract text and OCR only the inline images, not the
entire page. So that requires the following settings
Content-Type: *application/pdf*
**X-Tika-PDFextractInlineImages: *true*
X-Tika-PDFOcrStrategy: *auto* or *no_ocr*
I was under the impression that ocr strategy was something to do with how Tika
extracted text in general (i.e. applied OCR to the inline images or not)
> Using any setting other than AUTO or NO_OCR for X-Tika-PDFOcrStrategy causes
> remarkable performance loss
> --------------------------------------------------------------------------------------------------------
>
> Key: TIKA-3431
> URL: https://issues.apache.org/jira/browse/TIKA-3431
> Project: Tika
> Issue Type: Bug
> Components: tika-server
> Affects Versions: 1.26
> Reporter: Sal
> Priority: Minor
> Attachments: wiki.pdf
>
>
> When processing PDF document to the local Tika server using PUT request to
> endpoint [http://localhost:9998/tika.] If the PDFOcrStrategy is set to
> anything other than AUTO or NO_OCR, this causes extreme slowdown in
> processing of the PDF file.
>
> It doesn't matter if the PDF document has inline images or not, the slowdown
> happens regardless.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
