app.py

import streamlit as st
from transformers import BertForQuestionAnswering, BertTokenizer
import torch
from io import BytesIO
import PyPDF2
import pandas as pd
import spacy
from spacy.matcher import Matcher

# Load Spacy Model
nlp = spacy.load("en_core_web_sm")

# Extract Text from PDF
def extract_text_from_pdf(uploaded_file):
    pdf_reader = PyPDF2.PdfReader(BytesIO(uploaded_file.read()))
    resume_text = ''
    for page in pdf_reader.pages:
        resume_text += page.extract_text()
    return resume_text

# Load BERT Model for QA
model = BertForQuestionAnswering.from_pretrained("bert-large-uncased-whole-word-masking-finetuned-squad")
tokenizer = BertTokenizer.from_pretrained("bert-large-uncased-whole-word-masking-finetuned-squad")

# Generate Answer from QA Model
def answer_question(question, context, model, tokenizer):
    inputs = tokenizer.encode_plus(
        question,
        context,
        add_special_tokens=True,
        return_tensors="pt",
        truncation="only_second",
        max_length=512,
    )
    outputs = model(**inputs, return_dict=True)
    answer_start_scores = outputs.start_logits
    answer_end_scores = outputs.end_logits
    answer_start = torch.argmax(answer_start_scores)
    answer_end = torch.argmax(answer_end_scores) + 1
    input_ids = inputs["input_ids"].tolist()[0]
    answer = tokenizer.convert_tokens_to_string(
        tokenizer.convert_ids_to_tokens(input_ids[answer_start:answer_end])
    )
    return answer

# Extract Keywords for Resume Improvement
def extract_keywords_for_sections(text):
    doc = nlp(text)
    skills_keywords = set()
    project_keywords = set()

    # Define patterns for skills and project ideas
    skill_patterns = [[{"POS": "NOUN", "OP": "+"}], [{"POS": "PROPN", "OP": "+"}]]
    project_patterns = [[{"POS": "VERB"}, {"POS": "NOUN", "OP": "+"}], [{"POS": "VERB"}, {"POS": "PROPN", "OP": "+"}]]

    matcher = Matcher(nlp.vocab)
    matcher.add("SKILLS", skill_patterns)
    matcher.add("PROJECTS", project_patterns)

    for match_id, start, end in matcher(doc):
        span = doc[start:end]
        if nlp.vocab.strings[match_id] == "SKILLS":
            skills_keywords.add(span.text)
        elif nlp.vocab.strings[match_id] == "PROJECTS":
            project_keywords.add(span.text)

    return skills_keywords, project_keywords

# Suggest Resume Improvements
def suggest_resume_improvements(resume_text, job_description):
    skills_keywords, project_keywords = extract_keywords_for_sections(job_description)
    missing_skills = [kw for kw in skills_keywords if kw.lower() not in resume_text.lower()]
    potential_projects = [f"Consider a project involving '{keyword}'." for keyword in project_keywords]

    skill_suggestions = [f"Consider highlighting your experience or skills related to '{keyword}'." for keyword in missing_skills[:5]]
    project_suggestions = potential_projects[:5]

    return skill_suggestions, project_suggestions

# Analyze Matches between Resume and Job Description
def analyze_matches(resume_text, job_description):
    resume_keywords = set(extract_keywords_for_sections(resume_text)[0])
    job_desc_keywords = set(extract_keywords_for_sections(job_description)[0])

    matches = resume_keywords & job_desc_keywords
    if matches:
        commentary = f"Your resume matches the following keywords from the job description: {', '.join(matches)}"
    else:
        commentary = "There are no direct keyword matches between your resume and the job description."
    
    return commentary

# Initialize session state to store the log of QA pairs and satisfaction responses
if 'qa_log' not in st.session_state:
    st.session_state.qa_log = []

# Streamlit App Interface
st.title('Resume Enhancement and Analysis App')

# Resume PDF upload
uploaded_file = st.file_uploader("Upload your resume (PDF format):", type='pdf')
resume_text = ''
if uploaded_file is not None:
    resume_text = extract_text_from_pdf(uploaded_file)
    st.write("Resume Text:")
    st.write(resume_text)

    # Question-Answer Functionality
    user_question = st.text_input("Ask a question based on your resume:")
    if user_question:
        answer = answer_question(user_question, resume_text, model, tokenizer)
        st.write("Answer:")
        st.write(answer)

        # Log the interaction
        st.session_state.qa_log.append({
            'Question': user_question,
            'Answer': answer,
            'Satisfaction': 'Pending'
        })

# Job Description Input for Resume Improvement
job_description = st.text_area("Input the job description here for resume improvement suggestions:")
if job_description:
    skill_suggestions, project_suggestions = suggest_resume_improvements(resume_text, job_description)
    
    st.write('Technical Skill Improvement Suggestions:')
    for suggestion in skill_suggestions:
        st.write(suggestion)

    st.write('Notable Project Ideas:')
    for suggestion in project_suggestions:
        st.write(suggestion)

    # Analyze Matches and Provide Commentary
    match_commentary = analyze_matches(resume_text, job_description)
    st.write("Match Commentary:")
    st.write(match_commentary)

# User Feedback and Interaction Log
if st.session_state.qa_log:
    st.write("Interaction Log:")
    for i, interaction in enumerate(st.session_state.qa_log):
        if interaction['Satisfaction'] == 'Pending':
            satisfaction = st.radio(f'Are you satisfied with the answer to: "{interaction["Question"]}"?', ('Yes', 'No'), key=f'satisfaction_{i}')
            st.session_state.qa_log[i]['Satisfaction'] = satisfaction
    
    log_df = pd.DataFrame(st.session_state.qa_log)
    st.dataframe(log_df)