Spaces:
Sleeping
Sleeping
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,62 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# Step 1: Install the necessary libraries
|
2 |
+
# (Only needed locally; Hugging Face Spaces handles dependencies via 'requirements.txt')
|
3 |
+
# !pip install streamlit spacy numpy
|
4 |
+
|
5 |
+
import streamlit as st
|
6 |
+
import spacy
|
7 |
+
import numpy as np
|
8 |
+
import json
|
9 |
+
from numpy.linalg import norm
|
10 |
+
|
11 |
+
# Step 2: Load the spaCy model
|
12 |
+
nlp = spacy.load("en_core_web_md")
|
13 |
+
|
14 |
+
# Step 3: Load the FAQ data (ensure faqs.json is in the same directory)
|
15 |
+
with open('faqs.json', 'r') as f:
|
16 |
+
faqs = json.load(f)
|
17 |
+
|
18 |
+
# Step 4: Flatten the FAQ structure and precompute vectors
|
19 |
+
faq_docs = []
|
20 |
+
for category, faq_list in faqs.items():
|
21 |
+
for faq in faq_list:
|
22 |
+
question = faq['question']
|
23 |
+
answer = faq['answer']
|
24 |
+
faq_vector = nlp(question).vector # Precompute the vector
|
25 |
+
faq_docs.append((question, answer, faq_vector)) # Store question, answer, and vector
|
26 |
+
|
27 |
+
# Step 5: Define the function to find the most relevant FAQs
|
28 |
+
def find_most_relevant_faq_optimized(query, faq_docs):
|
29 |
+
"""Find the top 3 most relevant FAQs based on semantic similarity."""
|
30 |
+
query_vector = nlp(query).vector
|
31 |
+
|
32 |
+
# Calculate cosine similarity between query and each FAQ
|
33 |
+
similarities = [
|
34 |
+
(question, answer, np.dot(query_vector, faq_vector) / (norm(query_vector) * norm(faq_vector)))
|
35 |
+
for question, answer, faq_vector in faq_docs
|
36 |
+
]
|
37 |
+
|
38 |
+
# Sort by similarity score (highest first)
|
39 |
+
similarities = sorted(similarities, key=lambda x: x[2], reverse=True)
|
40 |
+
|
41 |
+
return similarities[:3] # Return top 3 FAQs
|
42 |
+
|
43 |
+
# Step 6: Create the Streamlit UI
|
44 |
+
st.title("Smart FAQ Search - SARAS AI Institute")
|
45 |
+
st.markdown("### Find Answers to Your Questions Instantly")
|
46 |
+
|
47 |
+
# Text input for the user query
|
48 |
+
query = st.text_input("Enter your question here:")
|
49 |
+
|
50 |
+
if query:
|
51 |
+
# Find the most relevant FAQs
|
52 |
+
top_faqs = find_most_relevant_faq_optimized(query, faq_docs)
|
53 |
+
|
54 |
+
# Display the results
|
55 |
+
st.markdown("### Top Relevant FAQs:")
|
56 |
+
for i, (question, answer, score) in enumerate(top_faqs, 1):
|
57 |
+
st.write(f"**{i}. {question}**")
|
58 |
+
st.write(f"*Answer:* {answer}")
|
59 |
+
st.write(f"**Similarity Score:** {score:.2f}")
|
60 |
+
else:
|
61 |
+
st.write("Please enter a query to search for relevant FAQs.")
|
62 |
+
|