Details
-
Wish
-
Status: Closed
-
Major
-
Resolution: Fixed
-
0.7.3
-
None
-
Windows XP / Java JDK 1.6.0_15 / Tika 0.4 with PDFbox-0.7.3.jar and fontbox-0.1.0.jar embedded
Description
Windings characters issue
-------------------------
If filed this question first in Tika's wish list (tika-331) but Ken Krugler suggest it was a PDFBox issue.
I have PDF files that include some characters in Windings font.
Tika parser replaces them with some Unicode characters that have nothing to do with the original, and, in some cases, replaces them with alphabetic characters. That is normal regarding these characters codes inside Windings font, but when hands pictures are replaced by alphabetic characters like A, B, etc. that disturbs further lexical analysis.
Would it be possible to improve the parsing and remplace these characters with more accurate Unicode characters ?
(see http://www.alanwood.net/demos/wingdings.html for possible correspondences).
Attached files :
test1.pdf is a PDF file including Windings characters. Some are commonly used by people, others less fequently.
Parsing_result1.txt is the text file produced by Tika.
test2.pdf is another example with the same WORD source file converted into PDF with another tool, and Parsing_result2.txt is the Tika parsing result. Windings characters are translated into different Unicode characters than with the previous version.
Spacing issue
-------------
Look at lines 10 and 11 in test2.pdf.
Look at lines 11 and 12 in Tika parsing result (Parsing_result2.txt) :
ðLocalisation des zones de livraison et de stockage
ðLocalisation des zones dangereuses
There is no space between ð and Localisation (ð is the translation of Winding's "Rightwards white arrow" by Tika).
If you copy and paste lines 10 and 11 in test2.pdf into a Notepad Window, you get :
ð Localisation des zones de livraison et de stockage
ð Localisation des zones dangereuses
...with a space between ð and Localisation.
In my case, the missing space after Tika parsing result in considering "ðLocalisation" as a single word in following analysis.
Regards
Attachments
Attachments
Issue Links
- depends upon
-
PDFBOX-11 CID to Unicode mapping
- Closed