[
https://issues.apache.org/jira/browse/TIKA-3005?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16990324#comment-16990324
]
Tilman Hausherr commented on TIKA-3005:
---------------------------------------
What we'd need is a gibberish detector. That won't go positive on a math
formula or on "foreign" text.
Another idea, don't know if possible: instead of classic OCR, do OCR on the
glyphs individually and reconstruct the table. To see what I mean open the
resume4 file wih PDFBox PDFDebugger and look at
{{Root/Pages/Kids/[0]/Resources/Font/F1}}. Imagine we reconstruct the "Unicode
Character" column by OCRing the "Glyph" column.
> Unintelligible text content from PDF file
> -----------------------------------------
>
> Key: TIKA-3005
> URL: https://issues.apache.org/jira/browse/TIKA-3005
> Project: Tika
> Issue Type: Bug
> Affects Versions: 1.22
> Reporter: Jorge Spinsanti
> Priority: Major
> Attachments: file1.pdf, file2.pdf, file3.pdf, resume_4.pdf
>
>
> If I get text content from attachment, Tika doesn't fail but the content is
> unintelligible
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
