suggest page.extract_words() word sequence same as page.extract_text()

[rahxphoon]

I would like to use page.extract_words() words' position feature, found the extracted text sequence is not like page.extract_text()
since page.extract_text() is extracted per same line
how can I make page.extract_text() sequence same as page.extract_text()

[jsvine]

Hi @rahxphoon, and thanks for your interest in this library. Unfortunately, I'm having some difficulty understanding the question. Could you provide additional explanation and a specific example?

[rahxphoon]

@cobaltautomationdev
seems like it got worse after Changing use_text_flow to True
{
'bottom': 420.2881999999999,
'direction': 1,
'doctop': 410.43819999999994,
'text': '2',
'top': 410.43819999999994,
'upright': True,
'x0': 56.43830000000005,
'x1': 61.914900000000046
},
{
'bottom': 420.6381999999999,
'direction': 1,
'doctop': 410.78819999999996,
'text': 'DEMURRAGE - TRANSHIPMENT',
'top': 410.78819999999996,
'upright': True,
'x0': 69.80000000000004,
'x1': 220.20665000000005
},
{
'bottom': 433.3381999999999,
'direction': 1,
'doctop': 423.48819999999995,
'text': 'LOWBED',
'top': 423.48819999999995,
'upright': True,
'x0': 69.80000000000005,
'x1': 112.48990000000005
},
{
'bottom': 443.76719999999995,
'direction': 1,
'doctop': 435.6672,
'text': '(02/Dec/22)',
'top': 435.6672,
'upright': True,
'x0': 92.70000000000005,
'x1': 135.01440000000005
},
{
'bottom': 453.96719999999993,
'direction': 1,
'doctop': 445.86719999999997,
'text': '(TRLU6423563)',
'top': 445.86719999999997,
'upright': True,
'x0': 94.95180000000005,
'x1': 153.02070000000006
},
{
'bottom': 470.18819999999994,
'direction': 1,
'doctop': 460.3382,
'text': '100.00 DAY Z',
'top': 460.3382,
'upright': True,
'x0': 411.9383,
'x1': 560.7183500000001
},
{
'bottom': 470.68819999999994,
'direction': 1,
'doctop': 460.8382,
'text': '1 100.00',
'top': 460.8382,
'upright': True,
'x0': 319.6883,
'x1': 551.1596000000001
},
{
'bottom': 472.37899999999996,
'direction': 1,
'doctop': 461.62899999999996,
'text': '$',
'top': 461.62899999999996,
'upright': True,
'x0': 477.95000000000005,
'x1': 483.927
}

the extracted text is much more in disorder now, it extract the second(third) line before completing the first line...

[rahxphoon]

I think extract_text uses the PDF's underlying flow of characters as a guide for ordering and segmenting the words
I wish the sequence can also applied to extract_words (text field)

[cobaltautomationdev]

maybe you can try to extract this part as table? extract_table treat a line as a row.

[rahxphoon]

thanks, that will be considered in the future
may be I will try manual re-construct the extract_words() text sequence same as extract_text
since extract_text sequence is robust, but in the same time, I need the position from extract_words

[jsvine]

Hi @rahxphoon, can you provide a copy of the PDF?

In the meantime, you may have luck adjusting the y_tolerance parameter.