Spaces:
Sleeping
Sleeping
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,66 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
from transformers import AutoModel, AutoTokenizer
|
3 |
+
import torch
|
4 |
+
import torch.nn.functional as F
|
5 |
+
|
6 |
+
# Load embedding model and tokenizer
|
7 |
+
model_name = "Supabase/gte-small"
|
8 |
+
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
9 |
+
model = AutoModel.from_pretrained(model_name)
|
10 |
+
model.eval()
|
11 |
+
|
12 |
+
def get_embedding(text):
|
13 |
+
inputs = tokenizer(text, return_tensors="pt", truncation=True, padding=True)
|
14 |
+
with torch.no_grad():
|
15 |
+
output = model(**inputs)
|
16 |
+
return output.last_hidden_state[:, 0, :].squeeze() # Use CLS token embedding
|
17 |
+
|
18 |
+
def get_similarity_and_excerpt(query, paragraph1, paragraph2, paragraph3, threshold_weight):
|
19 |
+
paragraphs = [p for p in [paragraph1, paragraph2, paragraph3] if p.strip()]
|
20 |
+
|
21 |
+
if not query.strip() or not paragraphs:
|
22 |
+
return "Please provide both a query and at least one document paragraph."
|
23 |
+
|
24 |
+
query_embedding = get_embedding(query)
|
25 |
+
ranked_paragraphs = []
|
26 |
+
|
27 |
+
for paragraph in paragraphs:
|
28 |
+
para_embedding = get_embedding(paragraph)
|
29 |
+
similarity = F.cosine_similarity(query_embedding, para_embedding, dim=0).item()
|
30 |
+
|
31 |
+
# Highlight words using threshold
|
32 |
+
tokens = tokenizer.tokenize(paragraph)
|
33 |
+
threshold = max(0.02, threshold_weight)
|
34 |
+
highlighted_text = " ".join(f"<b>{token}</b>" if similarity > threshold else token for token in tokens)
|
35 |
+
highlighted_text = tokenizer.convert_tokens_to_string(highlighted_text.split())
|
36 |
+
|
37 |
+
ranked_paragraphs.append({"similarity": similarity, "highlighted_text": highlighted_text})
|
38 |
+
|
39 |
+
ranked_paragraphs.sort(key=lambda x: x["similarity"], reverse=True)
|
40 |
+
|
41 |
+
output_html = "<table border='1' style='width:100%; border-collapse: collapse;'>"
|
42 |
+
output_html += "<tr><th>Cosine Similarity</th><th>Highlighted Paragraph</th></tr>"
|
43 |
+
for item in ranked_paragraphs:
|
44 |
+
output_html += f"<tr><td>{round(item['similarity'], 4)}</td><td>{item['highlighted_text']}</td></tr>"
|
45 |
+
output_html += "</table>"
|
46 |
+
|
47 |
+
return output_html
|
48 |
+
|
49 |
+
interface = gr.Interface(
|
50 |
+
fn=get_similarity_and_excerpt,
|
51 |
+
inputs=[
|
52 |
+
gr.Textbox(label="Query", placeholder="Enter your search query..."),
|
53 |
+
gr.Textbox(label="Document Paragraph 1", placeholder="Enter a paragraph to match...", lines=4),
|
54 |
+
gr.Textbox(label="Document Paragraph 2 (optional)", placeholder="Enter another paragraph...", lines=4),
|
55 |
+
gr.Textbox(label="Document Paragraph 3 (optional)", placeholder="Enter another paragraph...", lines=4),
|
56 |
+
gr.Slider(minimum=0.02, maximum=0.5, value=0.1, step=0.01, label="Similarity Threshold")
|
57 |
+
],
|
58 |
+
outputs=[gr.HTML(label="Ranked Paragraphs")],
|
59 |
+
title="Embedding-Based Similarity Highlighting",
|
60 |
+
description="Uses cosine similarity with Supabase/gte-small embeddings to rank paragraphs and highlight relevant words.",
|
61 |
+
allow_flagging="never",
|
62 |
+
live=True
|
63 |
+
)
|
64 |
+
|
65 |
+
if __name__ == "__main__":
|
66 |
+
interface.launch()
|