Final Version

app.py

@@ -1,15 +1,49 @@
+"""
+
+    This class implments a Gradio interface to generate an audio summary 
+    from the abstract of a PDF article.
+
+"""
+
 import gradio as gr
+from gradio_pdf import PDF
+from pdf_processor import summarize_abstract
+from audio_processor import generate_audio
+
+# Given a PDF file path, this function returns an audio summary of the abstract
+def abstract_to_audio(pdf_path):
+    """ This function retuns the audio generated from the summary of the abstract of PDF file.
+    
+    Args: (pdf_path  : str)
+    """
+    summarized_abstract = summarize_abstract(pdf_path)
+    audio, sampling_rate = generate_audio(summarized_abstract)
+    return sampling_rate, audio.T
+
+
+with gr.Blocks() as abstract_audio:
+   gr.Markdown("""
+               # PDF abstract audio summarize
+               Create an audio summary of the Abstract of the uploaded article.""")
+   
+   with gr.Row():
+       pdf_input = PDF(label="PDF File ...", interactive=True)
+       audio_output = gr.Audio(label="Audio Summary ...")
+      
+   audio_abstract_button = gr.Button("Generate audio summary") 
+   audio_abstract_button.click(abstract_to_audio, inputs=pdf_input, outputs=audio_output)
 
-def greet(name):
-    return "Hello " + name + "!!"
-
-iface = gr.Interface(
-    fn=greet,
-    inputs="file",
-    outputs="audio",
-    live=True,
-    title="Audio Processor",
-    description="Process audio files and return the processed audio.",
-)
-if __name__ == "__main__":
-    iface.launch()
+   gr.Examples(
+            examples=["Article 11 Hidden Technical Debt in Machine Learning Systems.pdf"],
+            inputs=pdf_input,
+            outputs=audio_output,
+            fn=abstract_to_audio,
+            cache_examples=False,
+   )
+   with gr.Accordion("Information: "):
+       gr.Markdown("This application creates an audio summary of the Abstract of the uploaded article. \n"
+                   "Make sure that the uploaded article is in the expected format......\n" 
+                   "Please note that the summarization is implemented using facebook/bart-large-cnn with maxlength= 50 to provide a meaningful summary and not \n"
+                   "only the first sentence.")
+       
+abstract_audio.launch()

audio_processor.py

@@ -0,0 +1,18 @@
+import numpy as np
+import scipy
+from transformers import pipeline
+
+
+def generate_audio(one_sentence_summary):
+    """ Generate an audio from the summary of the abstract of PDF file."""
+    synthesiser = pipeline("text-to-speech", "suno/bark-small")
+    speech = synthesiser(one_sentence_summary, forward_params={"do_sample": True})
+    return speech["audio"], speech["sampling_rate"]
+
+def convert_to_16_bit_wav(data):
+    # Based on: https://docs.scipy.org/doc/scipy/reference/generated/scipy.io.wavfile.write.html
+    if data.dtype == np.float32:
+        data = data / np.abs(data).max()
+        data = data * 32767
+        data = data.astype(np.int16)
+    return data

flagged/log.csv

@@ -0,0 +1,4 @@
+name,output,flag,username,timestamp
+,,,,2023-12-01 08:46:01.539926
+,,,,2023-12-04 18:19:49.008896
+"{""path"":""flagged/PDF File.../9c658cc31a3d3aca9f3f/Article 11 Hidden Technical Debt in Machine Learning Systems.pdf"",""url"":""http://127.0.0.1:7860/file=/private/var/folders/ch/cnzx1kvd1d9_f94xydxjcr3m0000gq/T/gradio/1eb131a34fbb508a9dd8b646950c65901d6f1a5b/Article 11 Hidden Technical Debt in Machine Learning Systems.pdf"",""size"":165614,""orig_name"":""Article 11 Hidden Technical Debt in Machine Learning Systems.pdf"",""mime_type"":""""}",,,,2023-12-05 12:14:20.913438

gradio_cached_examples/10/log.csv

@@ -0,0 +1,2 @@
+Audio Summary,flag,username,timestamp
+"{""path"":""gradio_cached_examples/10/Audio Summary/b5d3212345fc9c05bd0b/audio.wav"",""url"":null,""size"":null,""orig_name"":""audio.wav"",""mime_type"":null}",,,2023-12-05 15:32:45.408395

gradio_cached_examples/5/log.csv

@@ -0,0 +1,2 @@
+Audio Summary,flag,username,timestamp
+"{""path"":""gradio_cached_examples/5/Audio Summary/0ecec320413aacb22ba4/audio.wav"",""url"":null,""size"":null,""orig_name"":""audio.wav"",""mime_type"":null}",,,2023-12-05 15:36:16.937479

gradio_cached_examples/9/log.csv

@@ -0,0 +1,2 @@
+Audio Summary,flag,username,timestamp
+"{""path"":""gradio_cached_examples/9/Audio Summary/f9814d212607b3ca1d85/audio.wav"",""url"":null,""size"":null,""orig_name"":""audio.wav"",""mime_type"":null}",,,2023-12-05 15:39:46.300924

pdf_exctraction.py

@@ -0,0 +1,86 @@
+# This class contains the code provided for extracting content from a PDF file
+
+import gradio as gr
+import PyPDF2
+import pdfplumber
+from pdfminer.high_level import extract_pages
+from pdfminer.layout import LTTextContainer, LTChar
+
+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)
+
+def read_pdf(pdf_path):
+
+    if pdf_path is None:
+        raise gr.Error("A PDF file must be specified!")
+    # create a PDF file object
+    pdf_file_obj = open(pdf_path, 'rb')
+    # create a PDF reader object
+    pdf_reader = PyPDF2.PdfReader(pdf_file_obj)
+
+    # 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)):
+        # Initialize the variables needed for the text extraction from the page
+        page_text = []
+        line_format = []
+        text_from_images = []
+        text_from_tables = []
+        page_content = []
+        table_extraction_flag= False
+        # Open the pdf file
+        pdf = pdfplumber.open(pdf_path)
+        # 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
+
+        # 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
+    pdf_file_obj.close()
+    return text_per_page

pdf_processor.py

@@ -0,0 +1,24 @@
+import gradio as gr
+from pdf_exctraction import read_pdf
+from transformers import pipeline
+
+# Extract the Abstract from the content of the document
+def extract_abstract(pdf_path):
+    text_from_pdf = read_pdf(pdf_path)
+    abstract_text = ""
+    for page_content in text_from_pdf.values():
+        if "Abstract\n" in page_content[0]:
+            index_of_abstract = page_content[0].index("Abstract\n")
+            if index_of_abstract < len(page_content[0]) - 1:
+                abstract_text = page_content[0][index_of_abstract + 1] 
+    if abstract_text == "":
+        raise gr.Error("The article does not contains an Abstract or it is not in the expected format")
+    return abstract_text   
+
+# Summarized the abstract extracted from PDF
+def summarize_abstract(pdf_path):
+    abstract = extract_abstract(pdf_path)
+    abstract = abstract.replace("\n", " ").replace("- ", "").replace("  ", " ")
+    summarizer = pipeline("summarization", model="facebook/bart-large-cnn")
+    result = summarizer(abstract, max_length=50, min_length=10, length_penalty=2.0, num_beams=4, early_stopping=True)
+    return result[0]['summary_text']