app.py

# https://huggingface.co/spaces/marcolorenzi98/AAI-projects

# -*- coding: utf-8 -*-
"""AbstracTalk.ipynb

Automatically generated by Colaboratory.

Original file is located at
    https://colab.research.google.com/drive/1SsbXdZC55VNVB3CVBntZ7ugyA3eqsVFp

#Assessment 3 Audio Processing and AI in Production
Part 2

What to Do: Create a Hugging Face Space and publish the code you generated in the previous notebook.

How to Do It: Create a comprehensive package with all required files to publish the app. Use Gradio to design the interface. In the interface, specify the app's name, provide a brief description, and mention that your app only accepts PDFs with abstracts. Include examples of working PDFs in the app. Upload your app to Hugging Face Space and ensure it remains accessible throughout the grading period.

What to Deliver: Upload a compressed folder with a .zip or .rar extension. The folder should contain all the files that you uploaded to your Hugging Face Space. Please ADD as first line of the app.py file the address of the Space running the app as a Python Comment (see the example below). The app should keep running in order to be tested at the moment of grading.

#Install and import
"""


#from IPython.display import Audio
from transformers import pipeline
import torch
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
import gradio as gr

import numpy as np

import os

"""# PDF Reader

## Libraries + Code
"""

# To read the PDF
import PyPDF2
# To analyze the PDF layout and extract text
from pdfminer.high_level import extract_pages, extract_text
from pdfminer.layout import LTTextContainer, LTChar, LTRect, LTFigure
# To extract text from tables in PDF
import pdfplumber
# To extract the images from the PDFs
from PIL import Image
from pdf2image import convert_from_path
# To perform OCR to extract text from images
import pytesseract
# To remove the additional created files
import os

# Create a function to extract text

def text_extraction(element):
    # Extracting the text from the in-line text element
    line_text = element.get_text()

    # Find the formats of the text
    # Initialize the list with all the formats that appeared in the line of text
    line_formats = []
    for text_line in element:
        if isinstance(text_line, LTTextContainer):
            # Iterating through each character in the line of text
            for character in text_line:
                if isinstance(character, LTChar):
                    # Append the font name of the character
                    line_formats.append(character.fontname)
                    # Append the font size of the character
                    line_formats.append(character.size)
    # Find the unique font sizes and names in the line
    format_per_line = list(set(line_formats))

    # Return a tuple with the text in each line along with its format
    return (line_text, format_per_line)

# Create a function to crop the image elements from PDFs
def crop_image(element, pageObj):
    # Get the coordinates to crop the image from the PDF
    [image_left, image_top, image_right, image_bottom] = [element.x0,element.y0,element.x1,element.y1]
    # Crop the page using coordinates (left, bottom, right, top)
    pageObj.mediabox.lower_left = (image_left, image_bottom)
    pageObj.mediabox.upper_right = (image_right, image_top)
    # Save the cropped page to a new PDF
    cropped_pdf_writer = PyPDF2.PdfWriter()
    cropped_pdf_writer.add_page(pageObj)
    # Save the cropped PDF to a new file
    with open('cropped_image.pdf', 'wb') as cropped_pdf_file:
        cropped_pdf_writer.write(cropped_pdf_file)

# Create a function to convert the PDF to images
def convert_to_images(input_file,):
    images = convert_from_path(input_file)
    image = images[0]
    output_file = "PDF_image.png"
    image.save(output_file, "PNG")

# Create a function to read text from images
def image_to_text(image_path):
    # Read the image
    img = Image.open(image_path)
    # Extract the text from the image
    text = pytesseract.image_to_string(img)
    return text

# Extracting tables from the page

def extract_table(pdf_path, page_num, table_num):
    # Open the pdf file
    pdf = pdfplumber.open(pdf_path)
    # Find the examined page
    table_page = pdf.pages[page_num]
    # Extract the appropriate table
    table = table_page.extract_tables()[table_num]
    return table

# Convert table into the appropriate format
def table_converter(table):
    table_string = ''
    # Iterate through each row of the table
    for row_num in range(len(table)):
        row = table[row_num]
        # Remove the line breaker from the wrapped texts
        cleaned_row = [item.replace('\n', ' ') if item is not None and '\n' in item else 'None' if item is None else item for item in row]
        # Convert the table into a string
        table_string+=('|'+'|'.join(cleaned_row)+'|'+'\n')
    # Removing the last line break
    table_string = table_string[:-1]
    return table_string

def read_pdf(pdf_path):
  # create a PDF file object
  pdfFileObj = open(pdf_path, 'rb')
  # create a PDF reader object
  pdfReaded = PyPDF2.PdfReader(pdfFileObj)

  # Create the dictionary to extract text from each image
  text_per_page = {}
  # We extract the pages from the PDF
  for pagenum, page in enumerate(extract_pages(pdf_path)):
      print("Elaborating Page_" +str(pagenum))
      # Initialize the variables needed for the text extraction from the page
      pageObj = pdfReaded.pages[pagenum]
      page_text = []
      line_format = []
      text_from_images = []
      text_from_tables = []
      page_content = []
      # Initialize the number of the examined tables
      table_num = 0
      first_element= True
      table_extraction_flag= False
      # Open the pdf file
      pdf = pdfplumber.open(pdf_path)
      # Find the examined page
      page_tables = pdf.pages[pagenum]
      # Find the number of tables on the page
      tables = page_tables.find_tables()


      # Find all the elements
      page_elements = [(element.y1, element) for element in page._objs]
      # Sort all the elements as they appear in the page
      page_elements.sort(key=lambda a: a[0], reverse=True)

      # Find the elements that composed a page
      for i,component in enumerate(page_elements):
          # Extract the position of the top side of the element in the PDF
          pos= component[0]
          # Extract the element of the page layout
          element = component[1]

          # Check if the element is a text element
          if isinstance(element, LTTextContainer):
              # Check if the text appeared in a table
              if table_extraction_flag == False:
                  # Use the function to extract the text and format for each text element
                  (line_text, format_per_line) = text_extraction(element)
                  # Append the text of each line to the page text
                  page_text.append(line_text)
                  # Append the format for each line containing text
                  line_format.append(format_per_line)
                  page_content.append(line_text)
              else:
                  # Omit the text that appeared in a table
                  pass

          # Check the elements for images
          if isinstance(element, LTFigure):
              # Crop the image from the PDF
              crop_image(element, pageObj)
              # Convert the cropped pdf to an image
              convert_to_images('cropped_image.pdf')
              # Extract the text from the image
              image_text = image_to_text('PDF_image.png')
              text_from_images.append(image_text)
              page_content.append(image_text)
              # Add a placeholder in the text and format lists
              page_text.append('image')
              line_format.append('image')

          # Check the elements for tables
          if isinstance(element, LTRect):
              # If the first rectangular element
              if first_element == True and (table_num+1) <= len(tables):
                  # Find the bounding box of the table
                  lower_side = page.bbox[3] - tables[table_num].bbox[3]
                  upper_side = element.y1
                  # Extract the information from the table
                  table = extract_table(pdf_path, pagenum, table_num)
                  # Convert the table information in structured string format
                  table_string = table_converter(table)
                  # Append the table string into a list
                  text_from_tables.append(table_string)
                  page_content.append(table_string)
                  # Set the flag as True to avoid the content again
                  table_extraction_flag = True
                  # Make it another element
                  first_element = False
                  # Add a placeholder in the text and format lists
                  page_text.append('table')
                  line_format.append('table')

                  # Check if we already extracted the tables from the page
                  if element.y0 >= lower_side and element.y1 <= upper_side:
                      pass
                  elif not isinstance(page_elements[i+1][1], LTRect):
                      table_extraction_flag = False
                      first_element = True
                      table_num+=1


      # Create the key of the dictionary
      dctkey = 'Page_'+str(pagenum)
      # Add the list of list as the value of the page key
      text_per_page[dctkey]= [page_text, line_format, text_from_images,text_from_tables, page_content]

  # Closing the pdf file object
  pdfFileObj.close()

  try:

    # Deleting the additional files created
    os.remove('cropped_image.pdf')
    os.remove('PDF_image.png')
  finally:
    return text_per_page

"""#Functions

##Extract abstract
"""

def upload_file(files):
    file_paths = [file.name for file in files]
    return file_paths

def extract_abstract(path):

  text_per_page = read_pdf(path)

  abstract_found = False
  abstract_content = ""
  abstract_lenght = 700
  start_collecting = False

  for num_page in text_per_page:
    page_i = text_per_page[num_page][0]

    for index, word in enumerate(page_i):
      if ("abstract" in word.lower() or "summary" in word.lower()):
        abstract_found = True
        start_collecting = True
        continue

      if start_collecting:
        abstract_content += word + ' '
        # Check if the collected content contains "Introduction" to stop collecting
        if "introduction" in word.lower():
          break

    cleaned_abstract = ' '.join(abstract_content.splitlines()).replace('\n', ' ').replace('  ', ' ')


    if abstract_found:
      print("Abstract found")
      return cleaned_abstract
    else:
      print("Abstract not found")

def summarize_abstract(path):

  abstract_article = extract_abstract(path)

  INSTRUCTION = "summarize, simplify, and contextualize in one sentence: "
  tokenizer = AutoTokenizer.from_pretrained("haining/scientific_abstract_simplification")
  model = AutoModelForSeq2SeqLM.from_pretrained("haining/scientific_abstract_simplification")
  input_text = abstract_article
  encoding = tokenizer(INSTRUCTION + input_text,
                      max_length=672,
                      padding='max_length',
                      truncation=True,
                      return_tensors='pt')

  decoded_ids = model.generate(input_ids=encoding['input_ids'],
                               attention_mask=encoding['attention_mask'],
                               max_length=512,
                               top_p=.9,
                               do_sample=True)

  summary=tokenizer.decode(decoded_ids[0], skip_special_tokens=True)

  # Extract and print only the first sentence
  first_sentence = summary.split('.')[0] + '.'
  print(first_sentence)
  return first_sentence

def text_to_speech(sentence):

  #sentence = summarize_abstract (path)

  synthesiser = pipeline("text-to-speech", "suno/bark-small")

  speech = synthesiser(sentence, forward_params={"do_sample": True})

  audio_float32 = speech["audio"]
  sr = speech["sampling_rate"]

  #gr.Audio only accept a tuple(int, np.array(int16))
  audio_int16 = (audio_float32 * 32767).astype(np.int16)
  audio_reshaped = audio_int16.reshape(audio_int16.shape[1])

  return sr, audio_reshaped

def sum_audio(path):

  sentence = summarize_abstract (path)

  synthesiser = pipeline("text-to-speech", "suno/bark-small")

  speech = synthesiser(sentence, forward_params={"do_sample": True})

  audio_float32 = speech["audio"]
  sr = speech["sampling_rate"]

  #gr.Audio only accept a tuple(int, np.array(int16))
  audio_int16 = (audio_float32 * 32767).astype(np.int16)
  audio_reshaped = audio_int16.reshape(audio_int16.shape[1])

  audio_tuple = (sr, audio_reshaped)

  return sentence, audio_tuple

"""# Uploading PDF File"""

#from google.colab import files
#uploaded = files.upload()


"""#Gradio interface"""

interface = gr.Blocks()


with interface:
  gr.Markdown(
    """
    # AbstracTalk
    This app let's you upload an article (you can only upload a PDF with an abstract).
    It reads the abstract and does not only summarize it in just one sentence,
    but also makes it simpler for anybody to understand. Moreover, it also provides
    an additional layer of accessibility through spoken versions of the text.
    If you are not satisfied with the given summary you can press again the button and have a new summary.
    Have fun and master knowledge with AbstracTalk!
    """)

  #the interface architecture goes down here
  with gr.Row():
    with gr.Column():
      uploaded_article = gr.File()
      
    with gr.Column():
      summarized_abstract = gr.Textbox("One-sentence Abstract")
      talked_abstract = gr.Audio(type="numpy")
      with gr.Row():
        summary_button = gr.Button(value="Summarize Abstract", size="lg")
        tts_button = gr.Button(value="Speak Abstract", size="lg")
  
  gr.Markdown("## PDF Examples")
  gr.Examples(
      examples=[os.path.join(os.path.abspath(""), "Article 11 Hidden Technical Debt in Machine Learning Systems.pdf")],
      inputs=uploaded_article,
      outputs=[summarized_abstract, talked_abstract],
      fn=sum_audio,
      cache_examples = True,
    )

  #the functionality goes down here

  #first column


  #second column
  summary_button.click(summarize_abstract, inputs=uploaded_article, outputs=summarized_abstract)
  tts_button.click(text_to_speech, inputs=summarized_abstract, outputs=talked_abstract)

if __name__ == "__main__":
    interface.launch(debug=False)