Use lxml.etree, iterate ocr_line > ocr_word #57
Conversation
| except: | ||
| continue | ||
| for word in line.xpath('.//*[@class="ocrx_word"]'): | ||
| rawtext = word.xpath('./text()')[0] |
There was a problem hiding this comment.
This does only look at the first direct text node, but ignores more nested structure as e.g.
<span class='ocrx_word' id='word_1_1' title='bbox 118 3884 122 5088; x_wconf 95' lang='eng' dir='ltr'><strong><em> </em></strong></span>cf. #33 (comment)
I guess that we should go through all text nodes .//text() and concatenate the output. What do you think?
There was a problem hiding this comment.
Probably should parse as (X)HTML and use text_content.
| continue | ||
| for word in line.xpath('.//*[@class="ocrx_word"]'): | ||
| rawtext = word.xpath('./text()')[0] | ||
| # sys.stderr.write("WORD: '%s', type '%s'\n" % (rawtext, type(rawtext))) |
There was a problem hiding this comment.
I would suggest also to check afterwards rawtext and if it is None or empty up to some spaces, then we should continue. Or is there any reason to draw a word box with spaces only in the pdf?
There was a problem hiding this comment.
https://github.com/tmbdev/hocr- tools/pull/57/commits/fb994c30bce0df838506bf1d85c9f7dbf66e3928 should be sensible solution.
|
Thank you very much! |
[Move from https://github.com/UB-Mannheim/pull/19 here. ]
@kba writes:
This should not change existing behavior, but additionally allow processing non-XHTML-namespaced non-
spanocr_line.In the long run, integrating the code from jbarlow83/OCRmyPDF would be useful. Making the assumptions hocr-pdf makes on the hocr explicit would help too.