# -*- coding: utf-8 -*-
"""
Created on Mon Dec 25 18:18:27 2023
@author: danielshirazi
"""
import gradio as gr
import fitz # PyMuPDF
import questiongenerator as qs
import random
from sentence_transformers import SentenceTransformer, util
from questiongenerator import QuestionGenerator
qg = QuestionGenerator()
def highlight_similar_sentence(text1, text2, color='yellow'):
# Load the pre-trained sentence-transformers model
model = SentenceTransformer("paraphrase-MiniLM-L6-v2")
# Split text into sentences
sentences_text1 = [sentence.strip() for sentence in text1.split('.') if sentence.strip()]
sentences_text2 = [sentence.strip() for sentence in text2.split('.') if sentence.strip()]
# Compute embeddings for text1
#embeddings_text1 = model.encode(sentences_text1, convert_to_tensor=True)
highlighted_text2 = text2
max_similarity = 0.0
# Find the most similar sentence in text2 for each sentence in text1
for sentence_text1 in sentences_text1:
# Compute embeddings for the current sentence in text1
embedding_text1 = model.encode(sentence_text1, convert_to_tensor=True)
for sentence_text2 in sentences_text2:
# Compute cosine similarity between sentence in text1 and text2
embedding_text2 = model.encode(sentence_text2, convert_to_tensor=True)
similarity = util.pytorch_cos_sim(embedding_text1, embedding_text2).item()
# Highlight the most similar sentence in text2
if similarity > max_similarity:
max_similarity = similarity
highlighted_text2= highlight_text(text2, sentence_text2, color=color)
#highlighted_text2 = text2.replace(sentence_text2, f"<span style='background-color: {color};'>{sentence_text2}</span>")
return highlighted_text2
def Extract_QA(qlist,selected_extracted_text):
Q_All=''
A_All=''
xs=['A','B','C','D']
h_colors=['yellow', 'red', 'DodgerBlue', 'Orange', 'Violet']
for i in range(len(qlist)):
question_i= qlist[i]['question']
Choices_ans= []
Choice_is_correct=[]
for j in range(4):
Choices_ans= Choices_ans+ [qlist[i]['answer'][j]['answer']]
Choice_is_correct= Choice_is_correct+ [qlist[i]['answer'][j]['correct']]
Q=f"""
Q_{i+1}: {question_i}
A. {Choices_ans[0]}
B. {Choices_ans[1]}
C. {Choices_ans[2]}
D. {Choices_ans[3]}
"""
result = [x for x, y in zip(xs, Choice_is_correct) if y ]
correct_answer= [x for x, y in zip(Choices_ans, Choice_is_correct) if y ]
A= f"""
<p>Answer_{i+1}: {result[0]} - {correct_answer[0]}<p>
"""
color= h_colors[i]
A_sen= f""" The correct answer is {correct_answer[0]}."""
A= highlight_text(input_text=A, selcted_text=correct_answer[0], color=color)
selected_extracted_text= highlight_similar_sentence(A_sen, selected_extracted_text, color=color)
Q_All= Q_All+Q
A_All=A_All+ A
return (Q_All,A_All,selected_extracted_text)
def extract_text_from_pdf(pdf_file_path):
# Read the PDF file
global extracted_text
text = []
with fitz.open(pdf_file_path) as doc:
for page in doc:
text.append(page.get_text())
extracted_text= '\n'.join(text)
extracted_text= get_sub_text(extracted_text)
return ("The pdf is uploaded Successfully from:"+ str(pdf_file_path))
qg = qs.QuestionGenerator()
def get_sub_text(TXT):
sub_texts= qg._split_into_segments(TXT)
if isinstance(sub_texts, list):
return sub_texts
else:
return [sub_texts]
def highlight_text(input_text, selcted_text, color='yellow'):
# Replace 'highlight' with <span> tags for highlighting
highlighted_text = input_text.replace(selcted_text, f'<span style="background-color: {color}">{selcted_text}</span>')
return highlighted_text
def pick_One_txt(sub_texts):
global selected_extracted_text
N= len(sub_texts)
if N==1:
selected_extracted_text= sub_texts[0]
return(selected_extracted_text)
# Generate a random number between low and high
random_number = random.uniform(0, N)
# Pick the integer part of the random number
random_number = int(random_number)
selected_extracted_text= sub_texts[random_number]
return(selected_extracted_text)
def pipeline(NoQs):
global Q,A
text= selected_extracted_text
qlist= qg.generate(text, num_questions=NoQs, answer_style="multiple_choice")
Q,A,highligthed_text= Extract_QA(qlist,text)
A= A + '\n'+highligthed_text
return (Q,A)
def ReurnAnswer():
return A
def GetQuestion(NoQs):
NoQs=int(NoQs)
pick_One_txt(extracted_text)
Q,A=pipeline(NoQs)
return Q
with gr.Blocks() as demo:
with gr.Row():
with gr.Column(scale=1):
with gr.Row():
gr.Image("PupQuizAI.png")
gr.Markdown(""" 🐶 **PupQuizAI** is an Artificial-Intelligence tool that streamlines the studying process. Simply input a text pdf that you need to study from. Then, PupQuiz will create 1-5 custom questions for you to study from each time you push 'Show Questions'.
""" )
input_file=gr.UploadButton(label='Select a file!', file_types=[".pdf"])
input_file.upload(extract_text_from_pdf, input_file)
#upload_btn = gr.Button(value="Upload the pdf File.")
Gen_Question = gr.Button(value="Show Questions")
Gen_Answer = gr.Button(value="Show Answers")
No_Qs= gr.Slider(minimum=1, maximum=5,value=3, step=1, label='Max # of Questions')
gr.Markdown(""" 🐶
**Instructions:**
* Start by selecting a 'pdf' text file you want to upload by clicking the "Select file" button. (PupQuiz currently only supports files that can have highlightable text)
* Select the number of questions you want generated from the "# of Questions" selector.
* Click "Show Questions"
* Then, if you want answers to the questions, select "Show Answers" """ )
#gr.Image("PupQuizAI.png")
with gr.Column(scale=2.0):
#file_stat= gr.Textbox(label="File Status")
question = gr.Textbox(label="Question(s)")
#Answer = gr.Textbox(label="Answer(s)")
Answer = gr.HTML(label="Answer(s)")
Gen_Question.click(GetQuestion, inputs=No_Qs, outputs=question, api_name="QuestioGenerator")
Gen_Answer.click(ReurnAnswer, inputs=None, outputs=Answer, api_name="QuestioGenerator")
demo.launch()