Python is widely used for analyzing the data but the data need not be in the required format always. In such cases, we convert that format (like PDF or JPG, etc.) to the text format, in order to analyze the data in a better way. Python offers many libraries to do this task. There are several ways of doing this, including using libraries like PyPDF2 in Python. The major disadvantage of using these libraries is the encoding scheme. PDF documents can come in a variety of encodings including UTF-8, ASCII, Unicode, etc. So, converting the PDF to text might result in the loss of data due to the encoding scheme. Let’s see how to read all the contents of a PDF file and store it in a text document using OCR. Firstly, we need to convert the pages of the PDF to images and then, use OCR (Optical Character Recognition) to read the content from the image and store it in a text file.
Required Installations:
pip3 install PIL pip3 install pytesseract pip3 install pdf2image sudo apt-get install tesseract-ocr
There are two parts to the program as follows:
Part #1 deals with converting the PDF into image files. Each page of the PDF is stored as an image file. The names of the images stored are: PDF page 1 -> page_1.jpg PDF page 2 -> page_2.jpg PDF page 3 -> page_3.jpg …. PDF page n -> page_n.jpg.
Part #2 deals with recognizing text from the image files and storing it into a text file. Here, we process the images and convert it into text. Once we have the text as a string variable, we can do any processing on the text. For example, in many PDFs, when a line is completed, but a particular word cannot be written entirely in the same line, a hyphen (‘-‘) is added, and the word is continued on the next line. For example –
This is some sample text but this parti- cular word could not be written in the same line.
Now for such words, a fundamental pre-processing is done to convert the hyphen and the new line into a full word. After all the pre-processing is done, this text is stored in a separate text file. To get the input PDF files used in the code, click d.pdf.
Below is the implementation:
Python3
# Requires Python 3.6 or higher due to f-strings# Import librariesimport platformfrom tempfile import TemporaryDirectoryfrom pathlib import Pathimport pytesseractfrom pdf2image import convert_from_pathfrom PIL import Imageif platform.system() == "Windows": # We may need to do some additional downloading and setup... # Windows needs a PyTesseract Download pytesseract.pytesseract.tesseract_cmd = ( r"C:\Program Files\Tesseract-OCR\tesseract.exe" ) # Windows also needs poppler_exe path_to_poppler_exe = Path(r"C:\.....") # Put our output files in a sane place... out_directory = Path(r"~\Desktop").expanduser()else: out_directory = Path("~").expanduser() # Path of the Input pdfPDF_file = Path(r"d.pdf")# Store all the pages of the PDF in a variableimage_file_list = []text_file = out_directory / Path("out_text.txt")def main(): ''' Main execution point of the program''' with TemporaryDirectory() as tempdir: # Create a temporary directory to hold our temporary images. """ Part #1 : Converting PDF to images """ if platform.system() == "Windows": pdf_pages = convert_from_path( PDF_file, 500, poppler_path=path_to_poppler_exe ) else: pdf_pages = convert_from_path(PDF_file, 500) # Read in the PDF file at 500 DPI # Iterate through all the pages stored above for page_enumeration, page in enumerate(pdf_pages, start=1): # enumerate() "counts" the pages for us. # Create a file name to store the image filename = f"{tempdir}\page_{page_enumeration:03}.jpg" # Declaring filename for each page of PDF as JPG # For each page, filename will be: # PDF page 1 -> page_001.jpg # PDF page 2 -> page_002.jpg # PDF page 3 -> page_003.jpg # .... # PDF page n -> page_00n.jpg # Save the image of the page in system page.save(filename, "JPEG") image_file_list.append(filename) """ Part #2 - Recognizing text from the images using OCR """ with open(text_file, "a") as output_file: # Open the file in append mode so that # All contents of all images are added to the same file # Iterate from 1 to total number of pages for image_file in image_file_list: # Set filename to recognize text from # Again, these files will be: # page_1.jpg # page_2.jpg # .... # page_n.jpg # Recognize the text as string in image using pytesserct text = str(((pytesseract.image_to_string(Image.open(image_file))))) # The recognized text is stored in variable text # Any string processing may be applied on text # Here, basic formatting has been done: # In many PDFs, at line ending, if a word can't # be written fully, a 'hyphen' is added. # The rest of the word is written in the next line # Eg: This is a sample text this word here GeeksF- # orGeeks is half on first line, remaining on next. # To remove this, we replace every '-\n' to ''. text = text.replace("-\n", "") # Finally, write the processed text to the file. output_file.write(text) # At the end of the with .. output_file block # the file is closed after writing all the text. # At the end of the with .. tempdir block, the # TemporaryDirectory() we're using gets removed! # End of main function! if __name__ == "__main__": # We only want to run this if it's directly executed! main() |
Output: Input PDF file: 
Output Text file: 
As we see, the pages of the PDF were converted to images. Then the images were read, and the content was written into a text file. Advantages of this method include:
- Avoiding text-based conversion because of the encoding scheme resulting in loss of data.
- Even handwritten content in PDF can be recognized due to the usage of OCR.
- Recognizing only particular pages of the PDF is also possible.
- Getting the text as a variable so that any amount of required pre-processing can be done.
Disadvantages of this method include:
- Disk storage is used to store the images in the local system. Although these images are tiny in size.
- Using OCR cannot guarantee 100% accuracy. Given a computer-typed PDF document results in very high accuracy.
- Handwritten PDFs are still recognized, but the accuracy depends on various factors like handwriting, page color, etc.
