app.py

from PyPDF2 import PdfReader
from transformers import AutoModelForSeq2SeqLM, AutoTokenizer
from gtts import gTTS
import os

# Download the model and tokenizer
model_name = "ArtifactAI/led_large_16384_arxiv_summarization"
model = AutoModelForSeq2SeqLM.from_pretrained(model_name)
tokenizer = AutoTokenizer.from_pretrained(model_name)


def summarize_and_speak_pdf_abstract(pdf_path):
    """
    Reads a PDF file, extracts the abstract, summarizes it in one sentence, and generates an audio file of the summary.

    Args:
        pdf_path: Path to the PDF file.
    """

    # Summarize the abstract
    summary = summarize_pdf_abstract(pdf_path)

    # Define language and audio format
    language = "en"  # Change this to your desired language
    audio_format = "mp3"

    # Create the text-to-speech object
    tts = gTTS(text=summary, lang=language)

    # Generate the audio file
    audio_file_name = f"summary.{audio_format}"
    tts.save(audio_file_name)

    print(f"Audio file created: {audio_file_name}")

    # Play the audio file (optional)
    # os.system(f"play {audio_file_name}")


# Define the function to summarize the abstract
def summarize_pdf_abstract(pdf_path):
    """
    Reads a PDF file, extracts the abstract, and summarizes it in one sentence.

    Args:
        pdf_path: Path to the PDF file.

    Returns:
        A string containing the one-sentence summary of the abstract.
    """

    # Read the PDF file
    reader = PdfReader(open(pdf_path, "rb"))

    # Extract the abstract
    abstract_text = ""
    for page in reader.pages:
        # Search for keywords like "Abstract" or "Introduction"
        if (
            "Abstract" in page.extract_text()
            or "Introduction" in page.extract_text()
        ):
            # Extract the text following the keyword
            abstract_text = page.extract_text()
            break

    # Encode the abstract text
    inputs = tokenizer(abstract_text, return_tensors="pt")

    # Generate the summary
    outputs = model.generate(**inputs)

    # Decode the summary
    summary = tokenizer.decode(outputs[0], skip_special_tokens=True)

    return summary