incorrect_sentence_finder.py

from PyPDF2 import PdfReader
from openai import AzureOpenAI
import gradio as gr
import os
class IncorrectSentenceFinder:
    """
    This class finds and displays grammatically incorrect sentences in a PDF document using OpenAI's GPT-3.

    Args:
        pdf_file (str): The path to the PDF file.
    """

    def __init__(self):
        """
        Initialize the IncorrectSentenceFinder with the OpenAI API key.
        """
        
        # openai.api_type = os.getenv['api_type']
        # openai.api_base = os.getenv['api_base']
        # openai.api_version = os.getenv['api_version']
        # openai.api_key = os.getenv['api_key']
        pass

    def _find_incorrect_sentence(self, text: str) -> str:
        """
        Use OpenAI's GPT-3 to identify grammatically incorrect sentences in the given text.

        Args:
            text (str): Text to check for grammatical errors.

        Returns:
            str: Grammatically incorrect sentences identified by GPT-3.
        """
        # Create a request to OpenAI's GPT-3 engine to identify grammatically incorrect sentences.
        client = AzureOpenAI(api_key=os.getenv("AZURE_OPENAI_KEY"),  
                                api_version="2023-07-01-preview",
                                azure_endpoint = os.getenv("AZURE_OPENAI_ENDPOINT")
                                 )
            
        conversation = [
                {"role": "system", "content": "You are a helpful Error sentence finder."},
                {"role": "user", "content": f"""list out the grammatical error sentence in the given text:\n{text}
                 format:
                  error sentence:finded error sentence
                  """}
            ]
                    
                    # Call OpenAI GPT-3.5-turbo
        chat_completion = client.chat.completions.create(
            model = "GPT-3",
            messages = conversation,
            max_tokens=1000,
            temperature=0
        )
        response = chat_completion.choices[0].message.content
        return response 

    def get_incorrect_sentence(self, pdf_file: str) -> str:
        """
        Extract text from the PDF document and find grammatically incorrect sentences.

        Returns:
            str: Grammatically incorrect sentences identified by GPT-3.
        """
        try:
            # Open the PDF file using PyMuPDF's fitz library
            doc =PdfReader(pdf_file.name)
            incorrect_sentences = ''
            # Iterate through each page in the PDF document and extract the text
            for page_number in range(len(doc.pages)):
                            # Extract text from the page
                            page = doc.pages[page_number]
                            text = page.extract_text()
                            incorrect_sentences += self._find_incorrect_sentence(text)
            return incorrect_sentences

        except Exception as e:
            print(f"An error occurred: {str(e)}")
    def file_output_fnn(self,file_path):
        file_path = file_path.name
        return file_path      

    def gradio_interface(self):
        with gr.Blocks(css="style.css",theme='xiaobaiyuan/theme_brief') as demo:    
            with gr.Row(elem_id = "col-container",scale=0.80):
              with gr.Column(elem_id = "col-container",scale=0.80):
                file1 = gr.File(label="File",elem_classes="filenameshow")
            
              with gr.Column(elem_id = "col-container",scale=0.20):  
                upload_button1 = gr.UploadButton(
                    "Browse File",file_types=[".txt", ".pdf", ".doc", ".docx",".json",".csv"],
                    elem_classes="uploadbutton")
                incorrect_sentence = gr.Button("Get Headings",elem_classes="uploadbutton")
            
            with gr.Row(elem_id = "col-container",scale=0.60):    
                headings = gr.Textbox(label = "Headings")

        upload_button1.upload(self.file_output_fnn,upload_button1,file1)
        incorrect_sentence.click(self.get_incorrect_sentence,upload_button1,headings)