Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,111 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import streamlit as st
|
2 |
+
import numpy as np
|
3 |
+
from sentence_transformers import SentenceTransformer, util
|
4 |
+
|
5 |
+
# Initialize sentence transformer model
|
6 |
+
@st.cache_resource
|
7 |
+
def load_model():
|
8 |
+
return SentenceTransformer('all-MiniLM-L6-v2')
|
9 |
+
|
10 |
+
model = load_model()
|
11 |
+
|
12 |
+
def get_embedding_and_similarity(text1, text2):
|
13 |
+
# Get embeddings
|
14 |
+
embedding1 = model.encode(text1, convert_to_tensor=True)
|
15 |
+
embedding2 = model.encode(text2, convert_to_tensor=True)
|
16 |
+
|
17 |
+
# Calculate cosine similarity
|
18 |
+
similarity = util.pytorch_cos_sim(embedding1, embedding2).item()
|
19 |
+
return similarity
|
20 |
+
|
21 |
+
st.title("π€ Interactive Sentence Embeddings Explorer")
|
22 |
+
st.markdown("""
|
23 |
+
This demo helps you understand how sentence transformers work by comparing text similarities.
|
24 |
+
Try different sentences to see how the model captures semantic meaning!
|
25 |
+
""")
|
26 |
+
|
27 |
+
# Main comparison section
|
28 |
+
st.header("Compare Two Texts")
|
29 |
+
|
30 |
+
col1, col2 = st.columns(2)
|
31 |
+
with col1:
|
32 |
+
st.markdown("**First Text**")
|
33 |
+
text1 = st.text_area("Enter first text", height=100,
|
34 |
+
value="I love programming in Python")
|
35 |
+
|
36 |
+
with col2:
|
37 |
+
st.markdown("**Second Text**")
|
38 |
+
text2 = st.text_area("Enter second text", height=100,
|
39 |
+
value="Python is my favorite programming language")
|
40 |
+
|
41 |
+
if st.button("Calculate Similarity"):
|
42 |
+
similarity = get_embedding_and_similarity(text1, text2)
|
43 |
+
|
44 |
+
st.markdown("### Similarity Score")
|
45 |
+
st.progress(similarity)
|
46 |
+
st.write(f"Cosine Similarity: {similarity:.4f}")
|
47 |
+
|
48 |
+
if similarity > 0.8:
|
49 |
+
st.success("These texts are very similar!")
|
50 |
+
elif similarity > 0.5:
|
51 |
+
st.info("These texts are somewhat similar")
|
52 |
+
else:
|
53 |
+
st.warning("These texts are quite different")
|
54 |
+
|
55 |
+
# Interactive examples section
|
56 |
+
st.header("Try These Examples")
|
57 |
+
st.markdown("Click on any example to see how similar sentences are handled by the model")
|
58 |
+
|
59 |
+
examples = {
|
60 |
+
"Similar Meaning, Different Words": {
|
61 |
+
"text1": "The cat is sleeping on the couch",
|
62 |
+
"text2": "A feline is resting on the sofa"
|
63 |
+
},
|
64 |
+
"Similar Words, Different Meaning": {
|
65 |
+
"text1": "The bank is by the river",
|
66 |
+
"text2": "I need to go to the bank for money"
|
67 |
+
},
|
68 |
+
"Technical Similarity": {
|
69 |
+
"text1": "Python is a programming language",
|
70 |
+
"text2": "Java is used for coding software"
|
71 |
+
},
|
72 |
+
"Opposite Meanings": {
|
73 |
+
"text1": "The stock market is going up",
|
74 |
+
"text2": "The stock market is going down"
|
75 |
+
}
|
76 |
+
}
|
77 |
+
|
78 |
+
selected_example = st.selectbox("Choose an example", list(examples.keys()))
|
79 |
+
|
80 |
+
if st.button("Try this example"):
|
81 |
+
example = examples[selected_example]
|
82 |
+
similarity = get_embedding_and_similarity(example["text1"], example["text2"])
|
83 |
+
|
84 |
+
st.markdown("### Example Texts")
|
85 |
+
st.write("Text 1:", example["text1"])
|
86 |
+
st.write("Text 2:", example["text2"])
|
87 |
+
|
88 |
+
st.markdown("### Similarity Score")
|
89 |
+
st.progress(similarity)
|
90 |
+
st.write(f"Cosine Similarity: {similarity:.4f}")
|
91 |
+
|
92 |
+
# Educational section
|
93 |
+
st.header("π How It Works")
|
94 |
+
st.markdown("""
|
95 |
+
1. **Text to Embeddings**: The model converts each text into a high-dimensional vector (embedding)
|
96 |
+
2. **Similarity Calculation**: Cosine similarity between vectors is calculated
|
97 |
+
3. **Score Interpretation**:
|
98 |
+
- 1.0 = Identical meaning
|
99 |
+
- >0.8 = Very similar
|
100 |
+
- >0.5 = Somewhat similar
|
101 |
+
- <0.5 = Different meanings
|
102 |
+
""")
|
103 |
+
|
104 |
+
# Advanced settings
|
105 |
+
with st.expander("π§ Advanced Settings"):
|
106 |
+
st.markdown("""
|
107 |
+
**Current Model**: all-MiniLM-L6-v2
|
108 |
+
- Embedding Size: 384 dimensions
|
109 |
+
- Optimized for semantic similarity tasks
|
110 |
+
- Fast and efficient for real-time applications
|
111 |
+
""")
|