Spaces:
Sleeping
Sleeping
| from transformers import pipeline, AutoModelForQuestionAnswering, AutoTokenizer | |
| import gradio as gr | |
| from PyPDF2 import PdfReader | |
| import re | |
| # Modell und Tokenizer laden | |
| model_name = "deepset/roberta-base-squad2" | |
| model = AutoModelForQuestionAnswering.from_pretrained(model_name) | |
| tokenizer = AutoTokenizer.from_pretrained(model_name) | |
| qa_pipeline = pipeline("question-answering", model=model, tokenizer=tokenizer) | |
| # Funktion zum Extrahieren und Bereinigen von Text aus PDF | |
| def extract_text_from_pdf(pdf_path): | |
| reader = PdfReader(pdf_path) | |
| text = "" | |
| for page in reader.pages: | |
| text += page.extract_text() | |
| return text | |
| def clean_text(text): | |
| text = re.sub(r'\s+', ' ', text) | |
| text = re.sub(r'[^\w\s.,-]', '', text) | |
| return text.strip() | |
| def split_text_into_paragraphs(text, max_length=500): | |
| paragraphs = text.split("\n") | |
| refined_paragraphs = [] | |
| temp = "" | |
| for para in paragraphs: | |
| if len(temp) + len(para) <= max_length: | |
| temp += " " + para | |
| else: | |
| refined_paragraphs.append(temp.strip()) | |
| temp = para | |
| if temp: | |
| refined_paragraphs.append(temp.strip()) | |
| return refined_paragraphs | |
| def find_relevant_parts(question, context_parts): | |
| keywords = question.split() | |
| relevant_parts = [ | |
| part for part in context_parts if any(keyword.lower() in part.lower() for keyword in keywords) | |
| ] | |
| return relevant_parts if relevant_parts else context_parts | |
| def validate_and_refine_answer(answer): | |
| if not answer or len(answer.split()) < 5: | |
| return "Die Antwort konnte nicht eindeutig aus dem Dokument ermittelt werden." | |
| invalid_phrases = ["bluetooth", "hand", "ke", "eingelegt"] | |
| for phrase in invalid_phrases: | |
| answer = answer.replace(phrase, "") | |
| return answer.capitalize().strip() | |
| def chatbot_response(pdf_path, question): | |
| context = extract_text_from_pdf(pdf_path) | |
| context = clean_text(context) | |
| context_parts = split_text_into_paragraphs(context) | |
| relevant_parts = find_relevant_parts(question, context_parts) | |
| answers = [] | |
| for part in relevant_parts: | |
| try: | |
| result = qa_pipeline(question=question, context=part) | |
| answers.append(result['answer']) | |
| except Exception: | |
| continue | |
| final_answer = validate_and_refine_answer(" ".join(answers).strip()) | |
| return final_answer | |
| # Gradio-Interface | |
| pdf_input = gr.File(label="PDF-Datei hochladen", type="filepath") | |
| question_input = gr.Textbox(label="Frage eingeben", placeholder="Stelle eine Frage zu dem PDF-Dokument") | |
| response_output = gr.Textbox(label="Antwort") | |
| interface = gr.Interface( | |
| fn=chatbot_response, | |
| inputs=[pdf_input, question_input], | |
| outputs=response_output, | |
| title="Verbesserte PDF-Fragebeantwortung", | |
| description="Lade eine PDF-Datei hoch und stelle Fragen zu ihrem Inhalt. Antworten basieren nur auf den PDF-Inhalten." | |
| ) | |
| if __name__ == "__main__": | |
| interface.launch() | |