Spaces:
Running
Running
Commit
•
e4a1f31
1
Parent(s):
40e834e
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,59 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
import fitz # PyMuPDF for parsing PDF
|
3 |
+
import streamlit as st
|
4 |
+
from sentence_transformers import SentenceTransformer, util
|
5 |
+
|
6 |
+
# Load a pre-trained SentenceTransformer model
|
7 |
+
model_name = "paraphrase-MiniLM-L6-v2" # You can change this to a different model if needed
|
8 |
+
model = SentenceTransformer(model_name)
|
9 |
+
|
10 |
+
# Function to extract text from a PDF file
|
11 |
+
def extract_text_from_pdf(pdf_path):
|
12 |
+
text = ""
|
13 |
+
with fitz.open(pdf_path) as pdf_document:
|
14 |
+
for page_num in range(pdf_document.page_count):
|
15 |
+
page = pdf_document.load_page(page_num)
|
16 |
+
text += page.get_text()
|
17 |
+
return text
|
18 |
+
|
19 |
+
# Function to perform semantic search
|
20 |
+
def semantic_search(query, documents, top_k=5):
|
21 |
+
query_embedding = model.encode(query, convert_to_tensor=True)
|
22 |
+
|
23 |
+
# Convert the list of documents to embeddings
|
24 |
+
document_embeddings = model.encode(documents, convert_to_tensor=True)
|
25 |
+
|
26 |
+
# Compute cosine similarity scores of query with documents
|
27 |
+
cosine_scores = util.pytorch_cos_sim(query_embedding, document_embeddings)
|
28 |
+
|
29 |
+
# Sort the results in decreasing order
|
30 |
+
results = []
|
31 |
+
for idx in range(len(cosine_scores)):
|
32 |
+
results.append((documents[idx], cosine_scores[idx].item()))
|
33 |
+
results = sorted(results, key=lambda x: x[1], reverse=True)
|
34 |
+
|
35 |
+
return results[:top_k]
|
36 |
+
|
37 |
+
def main():
|
38 |
+
st.title("Semantic Search on PDF Documents")
|
39 |
+
|
40 |
+
query = st.text_input("Enter your query:")
|
41 |
+
pdf_file = st.file_uploader("Upload a PDF file:", type=["pdf"])
|
42 |
+
|
43 |
+
if st.button("Search"):
|
44 |
+
if pdf_file:
|
45 |
+
pdf_path = os.path.join("uploads", pdf_file.name)
|
46 |
+
with open(pdf_path, "wb") as f:
|
47 |
+
f.write(pdf_file.read())
|
48 |
+
|
49 |
+
pdf_text = extract_text_from_pdf(pdf_path)
|
50 |
+
search_results = semantic_search(query, [pdf_text])
|
51 |
+
os.remove(pdf_path) # Delete the uploaded file after processing
|
52 |
+
|
53 |
+
st.write(f"Search results for query: '{query}'")
|
54 |
+
for i, (result, score) in enumerate(search_results, start=1):
|
55 |
+
st.write(f"{i}. Score: {score:.2f}")
|
56 |
+
st.write(result)
|
57 |
+
|
58 |
+
if __name__ == "__main__":
|
59 |
+
main()
|