Extract and Sort Numbers in a PDF?

No. The only reliable way to extract data from PDF is to manually type it in. PDF is designed for printing and nothing else.

KooKilla,

Here's the Workflow:

The Run Shell Script Action is:

textutil -stdout -convert txt "$1" | grep -E  '[0-9]{1,}' > ~/Desktop/Report.txt
while read line
do
    name=$line
    s=$( echo "$name" | grep -Eo  '[0-9]{1,}' )
    a=($s)
    for i in "${a[@]}"
    do
         printf "%s %s\n" "$i": "$line" >> ~/Desktop/Report2.txt
    done
done < ~/Desktop/Report.txt
sort -n ~/Desktop/Report2.txt > ~/Desktop/ReportSorted.txt
rm "$1" ~/Desktop/Report.txt ~/Desktop/Report2.txt

The Workflow is:

So, if the PDF is:

"I am 20. My friend is 30. Her husband is 40."

"I am 30. My friend is 40. Her husband is 50."

ReportSorted.txt will be:

20: "I am 20. My friend is 30. Her husband is 40."

30: "I am 20. My friend is 30. Her husband is 40."

30: "I am 30. My friend is 40. Her husband is 50."

40: "I am 20. My friend is 30. Her husband is 40."

40: "I am 30. My friend is 40. Her husband is 50."

50: "I am 30. My friend is 40. Her husband is 50."

...and as etresoft said:

Manually verify with the PDF to make sure you haven't missed anything

KooKilla wrote:

Well I converted it to RTF. Is there any way I can do it now? Maybe I wasn't clear enough

It would be best to use a plain text format for such things. However, you will have to manually verify with the PDF to make sure you haven't missed anything. If you see text like "20" in the PDF, it may not be "20" after extracting the text. It may look fine in most places, but it could get corrupted in others. If you try it on another PDF, you might not get anything.

You're right about PDF's. When I extracted to plain text with Automator, it was full of hidden characters that grep choked on. I worked around this by having Automator extract to rtf, then I used textutil to convert from rtf to txt to clean-up the hidden characters.

Hello

You may also try something like the shell script below.

It will extract text per sentence not per line. However, its definition of sentence is so incomplete that it cannot extract complex sentence such as quoted sentence in sentence correctly. If it is real problem, we'd need to refine parsing logic.

Hope this may help,

H

#!/bin/bash

infile=~/desktop/a.pdf            # input pdf file
outfile=~/desktop/a.out.txt        # output text file

/System/Library/Frameworks/Ruby.framework/Versions/1.8/usr/bin/ruby -Ku <<'EOF' - "$infile" > "$outfile"
require 'osx/cocoa'
include OSX
require_framework 'PDFKit'

url = NSURL.fileURLWithPath(ARGV[0])
s = PDFDocument.alloc.initWithURL(url).string.to_s

re_sstop =                                                # re - sentence stop
    /
        (?: 
            [.?!]+            # 1* sentence punctuations
            [ ]*            #  * space
            (?:                # *1 closing quotations            
                ["'”’»]
            )?
        )
        |
        \z                    # end of string
    /xomu
re_delim = /(?: \s+ | \z )/xomu                            # re - token delimiter
re_blankline = /(?: \r\n | \r | \n ){2,}/xomu            # re - blank lines
re_sentence = /( .+? #{re_sstop} ) #{re_delim}/xomu        # re - sentence    

hash = {}
s.gsub!(re_blankline, ' ')    # remove blank lines that OSX::PDFDocument#string method inserts between pages
s.scan(re_sentence).each do |q|
    q.first.scan(/[0-9]+/).each do |n|
        n = n.to_i
        if hash.key?(n)
            hash[n] << q unless hash[n].include?(q)
        else
            hash[n] = [q]
        end
    end
end
hash.keys.sort.each do |k|
    print "# Category = #{k}\n\n"
    hash[k].each do |v|
        print "# Statement = \n#{v}\n\n"
    end
    print "-" * 70, "\n"
end
EOF

Hiroto,

Did something in my script not work?

Did what the OP needed when I tested it.

Hiroto,

If the PDF is:

"I am 20. My friend is 30. Her husband is 40."

"I am 30. My friend is 40. Her husband is 50."

OP Needs is looking for the following output (see: https://discussions.apple.com/thread/6012407#25232161 )

20: "I am 20. My friend is 30. Her husband is 40."

30: "I am 20. My friend is 30. Her husband is 40."

30: "I am 30. My friend is 40. Her husband is 50."

40: "I am 20. My friend is 30. Her husband is 40."

40: "I am 30. My friend is 40. Her husband is 50."

50: "I am 30. My friend is 40. Her husband is 50."

Not sure if your script is what's needed (...or maybe it is, need OP input on this)

Hello

An answer is prescribed for the question asked and precision of answer is constrained by precision of question.

Since the question is unclear about the origin of source text, I assumed that a) the source text is general article consisting of chapters of sections of paragraphs of sentences and b) the "fact" to be retrived is statement represented by sentence. And since the question does not specify the output format, I used my discretion in defining it. However, should the retrieved data be used in spreadsheet or database, csv or xml would have been better.

Regards,

H

PS. Given the source text in pdf:

I am 20. My friend is 30. Her husband is 40.
I am 30. My friend is 40. Her husband is 50.

my script will yield the following output:

# Category = 20

# Statement = 
I am 20.

----------------------------------------------------------------------
# Category = 30

# Statement = 
My friend is 30.

# Statement = 
I am 30.

----------------------------------------------------------------------
# Category = 40

# Statement = 
Her husband is 40.

# Statement = 
My friend is 40.

----------------------------------------------------------------------
# Category = 50

# Statement = 
Her husband is 50.

----------------------------------------------------------------------

A Bash, Ruby & Cocoa trifle!

That is a lovely, and admirable, piece of work! 🙂

Hiroto wrote:

An answer is prescribed for the question asked and precision of answer is constrained by precision of question.

The question in this post did not link to the prior post that did give this information. (However, I did link to that post in my 1st reply in this thread which did clear up the points that you mention.)

And since the question does not specify the output format, I used my discretion in defining it. However, should the retrieved data be used in spreadsheet or database, csv or xml would have been better.

OP is looking to use "Excel or Numbers or something else to create a chart? Something that will take anything that has like the number 16 and put it into a 16 category with all other lines that include that number?"

As it appears that the OP needs all lines that include the selected number, then if the line has "20", then

20: "I am 20. My friend is 30. Her husband is 40."

appears preferable to

"

I am 20."

(but not 100% sure if thats what the OP meant)

But I only output a plain text file, so I was a bit lazy in that regard.

It should be an easy exersise for you to adapt your script to create the chart requested.

Hello

Unfortunately I had not read the linked thread which in fact appears to be the original thread.

So I'd guess the original text is a flat collection of statements, each on its own line, in educational psychology or something? Because if it's normal article in such field, extraction per paragraph would only result in too coarse data, I'm afraid.

This is another example to demonstrate how important the precise description of problem is in the first place. If a problem is well defined, it is almost solved. 😉

All the best,

H

Agreed. Also, the test data supplied by the OP was limited at best.