File size: 6,274 Bytes
73601bf |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 |
import streamlit as st
import google.generativeai as genai
import fitz # PyMuPDF for PDF text extraction
import spacy
from transformers import AutoTokenizer, AutoModelForTokenClassification, pipeline
from transformers import AutoModelForSeq2SeqLM
from docx import Document
import re
import dateparser
from datetime import datetime
import os
# Load SpaCy model
nlp_spacy = spacy.load('en_core_web_sm')
# Load Babelscape NER model
tokenizer_ner = AutoTokenizer.from_pretrained("Babelscape/wikineural-multilingual-ner")
model_ner = AutoModelForTokenClassification.from_pretrained("Babelscape/wikineural-multilingual-ner")
nlp_ner = pipeline('ner', model=model_ner, tokenizer=tokenizer_ner, aggregation_strategy="simple")
# Load GLinER model
gliner_tokenizer = AutoTokenizer.from_pretrained("DAMO-NLP-SG/gliner-large")
gliner_model = AutoModelForSeq2SeqLM.from_pretrained("DAMO-NLP-SG/gliner-large")
def extract_info_with_gliner(text, info_type):
input_text = f"Extract {info_type} from: {text}"
input_ids = gliner_tokenizer(input_text, return_tensors="pt").input_ids
outputs = gliner_model.generate(input_ids, max_length=100)
return gliner_tokenizer.decode(outputs[0], skip_special_tokens=True)
class EnhancedNERPipeline:
def __init__(self, nlp_spacy, nlp_ner, gliner_model, gliner_tokenizer):
self.nlp_spacy = nlp_spacy
self.nlp_ner = nlp_ner
self.gliner_model = gliner_model
self.gliner_tokenizer = gliner_tokenizer
def __call__(self, text):
# SpaCy processing
doc = self.nlp_spacy(text)
# Babelscape NER processing
ner_results = self.nlp_ner(text)
# GLinER processing
gliner_companies = extract_info_with_gliner(text, "company names")
gliner_experience = extract_info_with_gliner(text, "years of experience")
gliner_education = extract_info_with_gliner(text, "educational institutions")
# Combine results
combined_entities = doc.ents + tuple(ner_results)
# Add GLinER results as custom attributes
doc._.gliner_companies = gliner_companies.split(', ')
doc._.gliner_experience = gliner_experience
doc._.gliner_education = gliner_education.split(', ')
# Update doc.ents with combined results for other entity types
doc.ents = [ent for ent in combined_entities if ent.label_ not in ["ORG"]]
return doc
# Create the enhanced pipeline
enhanced_nlp = EnhancedNERPipeline(nlp_spacy, nlp_ner, gliner_model, gliner_tokenizer)
def extract_companies(doc):
gliner_companies = set(doc._.gliner_companies)
spacy_babelscape_companies = set([ent.text for ent in doc.ents if ent.label_ == "ORG"])
return list(gliner_companies.union(spacy_babelscape_companies))
def extract_experience(doc):
gliner_experience = int(re.search(r'\d+', doc._.gliner_experience).group()) if doc._.gliner_experience else 0
spacy_experience = max([datetime.now().year - date.year for ent in doc.ents if ent.label_ == "DATE" and (date := dateparser.parse(ent.text)) and date.year <= datetime.now().year] or [0])
return max(gliner_experience, spacy_experience)
def extract_education(doc):
gliner_education = set(doc._.gliner_education)
spacy_babelscape_education = set([ent.text for ent in doc.ents if ent.label_ == "ORG" and any(keyword in ent.text.lower() for keyword in ["university", "college", "institute", "school"])])
return list(gliner_education.union(spacy_babelscape_education))
def extract_text_from_pdf(file):
pdf = fitz.open(stream=file.read(), filetype="pdf")
text = ""
for page in pdf:
text += page.get_text()
return text
def extract_text_from_doc(file):
doc = Document(file)
return " ".join([paragraph.text for paragraph in doc.paragraphs])
def authenticate_gemini(api_key):
try:
genai.configure(api_key=api_key)
model = genai.GenerativeModel('gemini-pro')
return model
except Exception as e:
st.error(f"Authentication failed: {e}")
return None
def generate_summary(text, model):
prompt = f"Summarize the following resume:\n\n{text}\n\nProvide a brief overview of the candidate's qualifications, experience, and key skills."
response = model.generate_content(prompt)
return response.text
def main():
st.title("Enhanced Resume Analyzer with GLinER Focus")
api_key = st.text_input("Enter your Google Gemini API key", type="password")
uploaded_file = st.file_uploader("Choose a PDF or DOCX file", type=["pdf", "docx"])
if uploaded_file is not None and api_key:
try:
model = authenticate_gemini(api_key)
if model is None:
return
if uploaded_file.type == "application/pdf":
resume_text = extract_text_from_pdf(uploaded_file)
elif uploaded_file.type == "application/vnd.openxmlformats-officedocument.wordprocessingml.document":
resume_text = extract_text_from_doc(uploaded_file)
else:
st.error("Unsupported file format.")
return
# Process the resume text with the enhanced pipeline
doc = enhanced_nlp(resume_text)
companies = extract_companies(doc)
experience = extract_experience(doc)
education = extract_education(doc)
# Use GLinER for other extractions
phone = extract_info_with_gliner(resume_text, "phone number")
email = extract_info_with_gliner(resume_text, "email address")
linkedin = extract_info_with_gliner(resume_text, "LinkedIn profile")
st.subheader("Extracted Information")
st.write(f"**Years of Experience:** {experience}")
st.write("**Companies:**", ", ".join(companies))
st.write("**Education:**", ", ".join(education))
st.write(f"**Phone Number:** {phone}")
st.write(f"**Email:** {email}")
st.write(f"**LinkedIn:** {linkedin}")
summary = generate_summary(resume_text, model)
st.subheader("Resume Summary")
st.write(summary)
except Exception as e:
st.error(f"Error during processing: {e}")
if __name__ == "__main__":
main() |