Links extraction

[meghanaviyyapu]

I have tried extracting link text from the below PDF. Along with the link text was able to view additional text.
Have used page.get_textbox(rect) to extract link text. Could you let me know how only link text can be obtained
GUL_link.pdf

[JorjMcKie]

Your PDF creator unfortunately uses a too small inter-line distance. So line bboxes overlap each other:

Method page.get_textbox() is optimized for including any character that overlaps the desired rectangle. It therefore is not a good candidate here.
Using page.get_text(clip=link["from"]) is better but still includes "y" and "g" - the reason can be seen in the image.

for link in page.get_links():
    print(page.get_text(clip=link["from"]).replace("\n",""))

    
www.MetLife.com
ygHIPAAprivacyAmericasUS@metlife.com

You could experiment a little and reduce the height of the link rect a bit from its top:

for link in page.get_links():
    print(page.get_text(clip=link["from"]+(0,2,0,0)))

    
www.MetLife.com

HIPAAprivacyAmericasUS@metlife.com

This is much better.

[JorjMcKie]

Difficult / hard to predict! Because there so many sloppy PDF creators!
You could change your strategy somewhat:

1. make a list of words from the full page
2. when iteration over the link["from"] rectangles, select words that "mostly are contained in the link rectangle. Something like this:

# prepare list of word rectangles and word text
words=[(fitz.Rect(w[:4]), w[4]) for w in page.get_text("words")]

# iterate over the links and select words whose rectangles are inside by at last 80%:
for link in page.get_links():
    frect=link["from"]
    link_area = 0.8 * abs(frect)  # area of link rect
    for w, text in words:  # check word intersection size
        if abs(w & frect) >= link_area:
            print(text)

            
www.MetLife.com.
HIPAAprivacyAmericasUS@metlife.com,

[JorjMcKie]

To get rid of the trailing punctuations use punct = tuple(string.punctuation) and replace the text by text = text[:-1] if text.endswith(punct) else text.

[meghanaviyyapu]

Thanks a lot for the detailed examples

[meghanaviyyapu]

I have tried extracting links from the below PDF
BTA_link.pdf
The link got identified as "CERTIFICATE FACE PAGE"
When I tried extracting spans, it got extracted as CERTIFICATE PAG and E...........

Could you let me know if it is possible to get the span output identical to link output like CERTIFICATE PAGE instead of CERTIFICATE PAG?

[JorjMcKie]

The span looks like this:

pprint(spans[3])
{'ascender': 0.9052734375,
 'bbox': (72.0, 84.2972412109375, 560.7783203125, 95.4691162109375),
 'color': 0,
 'descender': -0.2119140625,
 'flags': 4,
 'font': 'ArialMT',
 'origin': (72.0, 93.3499755859375),
 'size': 10.0,
 'text': 'CERTIFICATE FACE PAGE............................................................................................................................... '
         '2 '}

And thus covers the full line.
But there are 3 different links covering this same line:

IAW for the text of the span you will not find a link that covers it completely.
So you somehow have to deal with this by adjusting your code appropriately ...

[meghanaviyyapu]

Thanks, is there a way to find out if the same span contains multiple links?

[JorjMcKie]

Look at the red rectangles! These are the "from" rectangles of the links.