PDF is very complex and I'm not specialist but I took source code of extractText() to see how it works and using print('>>>', operator, operands) I could see what values it found in PDF.
In this document it uses "Tm" to move position to new line so changed original code in extractText() and I used "Tm" to add
and I got text in lines
Arto Anttila. 1995. How to recognise subjects in
English. In Karlsson et al., chapt. 9, pp. 315-358.
Dekang Lin. 1996. Evaluation of Principar with the
Susanne corpus. In John Carroll, editor, Work-
shop on Robust Parsing, pages 54-69, Prague.
Jason M. Eisner. 1996. Three new probabilistic
models for dependency parsing: An exploration.
In The 16th International Conference on Compu-
tational Linguistics, pages 340-345. Copenhagen.
David G. Hays. 1964. Dependency theory: A
formalism and some observations. Language,
40(4):511-525.
Or with --- between lines
---
Arto Anttila. 1995. How to recognise subjects in
---
English. In Karlsson et al., chapt. 9, pp. 315-358.
---
Dekang Lin. 1996. Evaluation of Principar with the
---
Susanne corpus. In John Carroll, editor, Work-
---
shop on Robust Parsing, pages 54-69, Prague.
---
Jason M. Eisner. 1996. Three new probabilistic
---
models for dependency parsing: An exploration.
---
In The 16th International Conference on Compu-
---
tational Linguistics, pages 340-345. Copenhagen.
---
David G. Hays. 1964. Dependency theory: A
---
formalism and some observations. Language,
---
40(4):511-525.
But it still not so usefull but now code which I used to get this result
import PyPDF2
from PyPDF2.pdf import * # to import function used in origimal `extractText`
# --- functions ---
def myExtractText(self):
# code from original `extractText()`
# https://github.com/mstamy2/PyPDF2/blob/d7b8d3e0f471530267827511cdffaa2ab48bc1ad/PyPDF2/pdf.py#L2645
text = u_("")
content = self["/Contents"].getObject()
if not isinstance(content, ContentStream):
content = ContentStream(content, self.pdf)
for operands, operator in content.operations:
# used only for test to see values in variables
#print('>>>', operator, operands)
if operator == b_("Tj"):
_text = operands[0]
if isinstance(_text, TextStringObject):
text += _text
elif operator == b_("T*"):
text += "
"
elif operator == b_("'"):
text += "
"
_text = operands[0]
if isinstance(_text, TextStringObject):
text += operands[0]
elif operator == b_('"'):
_text = operands[2]
if isinstance(_text, TextStringObject):
text += "
"
text += _text
elif operator == b_("TJ"):
for i in operands[0]:
if isinstance(i, TextStringObject):
text += i
text += "
"
# new code to add `
` when text moves to new line
elif operator == b_("Tm"):
text += '
'
return text
# --- main ---
pdfFileObj = open('A97-1011.pdf', 'rb')
pdfReader = PyPDF2.PdfFileReader(pdfFileObj)
text = ''
for page in pdfReader.pages:
#text += page.extractText() # original function
text += myExtractText(page) # modified function
# get only text after word `References`
pos = text.lower().find('references')
text = text[pos+len('references '):]
# print all at once
print(text)
# print line by line
for line in text.split('
'):
print(line)
print('---')
After digging it seems Tm has also values and there is new postion x, y which I used to calculate distance between lines of text and I add
when distance is bigger then some value. I tested different values and from value 17 I got expected result
---
Arto Anttila. 1995. How to recognise subjects in English. In Karlsson et al., chapt. 9, pp. 315-358.
---
Dekang Lin. 1996. Evaluation of Principar with the Susanne corpus. In John Carroll, editor, Work- shop on Robust Parsing, pages 54-69, Prague.
---
Jason M. Eisner. 1996. Three new probabilistic models for dependency parsing: An exploration. In The 16th International Conference on Compu- tational Linguistics, pages 340-345. Copenhagen.
---
David G. Hays. 1964. Dependency theory: A formalism and some observations. Language, 40(4):511-525.
---
Here code
import PyPDF2
from PyPDF2.pdf import * # to import function used in origimal `extractText`
# --- functions ---
def myExtractText2(self):
# original code from `page.extractText()`
# https://github.com/mstamy2/PyPDF2/blob/d7b8d3e0f471530267827511cdffaa2ab48bc1ad/PyPDF2/pdf.py#L2645
text = u_("")
content = self["/Contents"].getObject()
if not isinstance(content, ContentStream):
content = ContentStream(content, self.pdf)
prev_x = 0
prev_y = 0
for operands, operator in content.operations:
# used only for test to see values in variables
#print('>>>', operator, operands)
if operator == b_("Tj"):
_text = operands[0]
if isinstance(_text, TextStringObject):
text += _text
elif operator == b_("T*"):
text += "
"
elif operator == b_("'"):
text += "
"
_text = operands[0]
if isinstance(_text, TextStringObject):
text += operands[0]
elif operator == b_('"'):
_text = operands[2]
if isinstance(_text, TextStringObject):
text += "
"
text += _text
elif operator == b_("TJ"):
for i in operands[0]:
if isinstance(i, TextStringObject):
text += i
text += "
"
elif operator == b_("Tm"):
x = operands[-2]
y = operands[-1]
diff_x = prev_x - x
diff_y = prev_y - y
#print('>>>', diff_x, diff_y - y)
#text += f'| {diff_x}, {diff_y - y} |'
if diff_y > 17 or diff_y < 0: # (bigger margin) or (move to top in next column)
text += '
'
#text += '
' # to add empty line between elements
prev_x = x
prev_y = y
return text
# --- main ---
pdfFileObj = open('A97-1011.pdf', 'rb')
pdfReader = PyPDF2.PdfFileReader(pdfFileObj)
text = ''
for page in pdfReader.pages:
#text += page.extractText() # original function
text += myExtractText(page) # modified function
# get only text after word `References`
pos = text.lower().find('references')
text = text[pos+len('references '):]
# print all at once
print(text)
# print line by line
for line in text.split('
'):
print(line)
print('---')
It works for this PDF but other files may have different structure or different distance between references and they may need other changes.
EDIT:
Little more universal version - it gets second argument
If you run without second argument
text += myExtractText(page)
then it works like original extractText() and you get all in one string.
If second argument is True
text += myExtractText(page, True)
then it adds new line after every Tm - like in my first version.
If second argument is integer number - ie. 17
text += myExtractText(page, 17)
then it adds new line when distance is bigger then 17 - like in my second version.
import PyPDF2
from PyPDF2.pdf import * # to import function used in origimal `extractText`
# --- functions ---
def myExtractText(self, distance=None):
# original code from `page.extractText()`
# https://github.com/mstamy2/PyPDF2/blob/d7b8d3e0f471530267827511cdffaa2ab48bc1ad/PyPDF2/pdf.py#L2645
text = u_("")
content = self["/Contents"].getObject()
if not isinstance(content, ContentStream):
content = ContentStream(content, self.pdf)
prev_x = 0
prev_y = 0
for operands, operator in content.operations:
# used only for test to see values in variables
#print('>>>', operator, operands)
if operator == b_("Tj"):
_text = operands[0]
if isinstance(_text, TextStringObject):
text += _text
elif operator == b_("T*"):
text += "
"
elif operator == b_("'"):
text += "
"
_text = operands[0]
if isinstance(_text, TextStringObject):
text += operands[0]
elif operator == b_('"'):
_text = operands[2]
if isinstance(_text, TextStringObject):
text += "
"
text += _text
elif operator == b_("TJ"):
for i in operands[0]:
if isinstance(i, TextStringObject):
text += i
text += "
"
if operator == b_("Tm"):
if distance is True:
text += '
'
elif isinstance(distance, int):
x = operands[-2]
y = operands[-1]
diff_x = prev_x - x
diff_y = prev_y - y
#print('>>>', diff_x, diff_y - y)
#text += f'| {diff_x}, {diff_y - y} |'
if diff_y > distance or diff_y < 0: # (bigger margin) or (move to top in next column)
text += '
'
#text += '
' # to add empty line between elements
prev_x = x
prev_y = y
return text
# --- main ---
pdfFileObj = open('A97-1011.pdf', 'rb')
pdfReader = PyPDF2.PdfFileReader(pdfFileObj)
text = ''
for page in pdfReader.pages:
#text += page.extractText() # original function
#text += myExtractText(page) # modified function (works like original version)
#text += myExtractText(page, True) # modified function (add `
` after every `Tm`)
text += myExtractText(page, 17) # modified function (add `
` only if distance is bigger then `17`)
# get only text after word `References`
pos = text.lower().find('references')
text = text[pos+len('references '):]
# print all at once
print(text)
# print line by line
for line in text.split('
'):
print(line)
print('---')
BTW: It can be useful not only for References but also for rest of text - It seems it split paragraphs.
Result for beginning of PDF
