Spaces:
Running
on
Zero
Running
on
Zero
Update app.py
Browse files
app.py
CHANGED
@@ -1,37 +1,46 @@
|
|
1 |
import spaces
|
2 |
import gradio as gr
|
3 |
from phi3_instruct_graph import MODEL_LIST, Phi3InstructGraph
|
4 |
-
from textwrap import dedent
|
5 |
import rapidjson
|
6 |
-
import spaces
|
7 |
from pyvis.network import Network
|
8 |
import networkx as nx
|
9 |
import spacy
|
10 |
from spacy import displacy
|
11 |
from spacy.tokens import Span
|
12 |
import random
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
13 |
|
14 |
-
|
15 |
-
def extract(text, model):
|
16 |
-
model = Phi3InstructGraph(model=model)
|
17 |
-
result = model.extract(text)
|
18 |
-
return rapidjson.loads(result)
|
19 |
-
|
20 |
def handle_text(text):
|
21 |
return " ".join(text.split())
|
22 |
|
23 |
-
|
24 |
-
|
25 |
-
|
26 |
-
|
27 |
-
|
28 |
-
|
29 |
-
|
30 |
-
|
31 |
-
|
32 |
-
|
33 |
-
def get_random_color():
|
34 |
-
return f"#{random.randint(0, 0xFFFFFF):06x}"
|
35 |
|
36 |
def find_token_indices(doc, substring, text):
|
37 |
result = []
|
@@ -48,9 +57,7 @@ def find_token_indices(doc, substring, text):
|
|
48 |
if token.idx + len(token) == end_index:
|
49 |
end_token = token.i + 1
|
50 |
|
51 |
-
if start_token is None
|
52 |
-
print(f"Token boundaries not found for '{substring}' at index {start_index}")
|
53 |
-
else:
|
54 |
result.append({
|
55 |
"start": start_token,
|
56 |
"end": end_token
|
@@ -59,12 +66,8 @@ def find_token_indices(doc, substring, text):
|
|
59 |
# Search for next occurrence
|
60 |
start_index = text.find(substring, end_index)
|
61 |
|
62 |
-
if not result:
|
63 |
-
print(f"Token boundaries not found for '{substring}'")
|
64 |
-
|
65 |
return result
|
66 |
|
67 |
-
|
68 |
def create_custom_entity_viz(data, full_text):
|
69 |
nlp = spacy.blank("xx")
|
70 |
doc = nlp(full_text)
|
@@ -82,8 +85,6 @@ def create_custom_entity_viz(data, full_text):
|
|
82 |
overlapping = any(s.start < end and start < s.end for s in spans)
|
83 |
if not overlapping:
|
84 |
span = Span(doc, start, end, label=node["type"])
|
85 |
-
|
86 |
-
# print(span)
|
87 |
spans.append(span)
|
88 |
if node["type"] not in colors:
|
89 |
colors[node["type"]] = get_random_light_color()
|
@@ -101,26 +102,28 @@ def create_custom_entity_viz(data, full_text):
|
|
101 |
html = displacy.render(doc, style="span", options=options)
|
102 |
return html
|
103 |
|
104 |
-
|
105 |
def create_graph(json_data):
|
106 |
G = nx.Graph()
|
107 |
|
|
|
108 |
for node in json_data['nodes']:
|
109 |
G.add_node(node['id'], title=f"{node['type']}: {node['detailed_type']}")
|
110 |
|
|
|
111 |
for edge in json_data['edges']:
|
112 |
G.add_edge(edge['from'], edge['to'], title=edge['label'], label=edge['label'])
|
113 |
|
|
|
114 |
nt = Network(
|
115 |
width="720px",
|
116 |
height="600px",
|
117 |
directed=True,
|
118 |
notebook=False,
|
119 |
-
bgcolor="#
|
120 |
-
font_color="
|
121 |
-
# bgcolor="#FFFFFF",
|
122 |
-
# font_color="#111827"
|
123 |
)
|
|
|
|
|
124 |
nt.from_nx(G)
|
125 |
nt.barnes_hut(
|
126 |
gravity=-3000,
|
@@ -130,71 +133,141 @@ def create_graph(json_data):
|
|
130 |
damping=0.09,
|
131 |
overlap=0,
|
132 |
)
|
133 |
-
|
134 |
# Customize edge appearance
|
135 |
-
|
136 |
-
|
137 |
-
|
138 |
-
|
139 |
-
|
140 |
-
|
141 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
142 |
html = nt.generate_html()
|
143 |
html = html.replace("'", '"')
|
144 |
|
145 |
-
return f"""<iframe style="width:
|
146 |
-
allow="midi; geolocation; microphone; camera; display-capture; encrypted-media;"
|
147 |
sandbox="allow-modals allow-forms allow-scripts allow-same-origin allow-popups
|
148 |
allow-top-navigation-by-user-activation allow-downloads" allowfullscreen=""
|
149 |
allowpaymentrequest="" frameborder="0" srcdoc='{html}'></iframe>"""
|
150 |
-
|
151 |
|
152 |
-
def process_and_visualize(text, model):
|
153 |
if not text or not model:
|
154 |
-
raise gr.Error("
|
|
|
|
|
155 |
json_data = extract(text, model)
|
|
|
|
|
156 |
entities_viz = create_custom_entity_viz(json_data, text)
|
157 |
|
|
|
158 |
graph_html = create_graph(json_data)
|
159 |
-
return graph_html, entities_viz, json_data
|
160 |
-
|
161 |
-
|
162 |
-
|
163 |
-
with gr.Blocks(title="Phi-3 Instruct Graph (by Emergent Methods") as demo:
|
164 |
-
gr.Markdown("# Phi-3 Instruct Graph (by Emergent Methods)")
|
165 |
-
gr.Markdown("Extract a JSON graph from a text input and visualize it.")
|
166 |
-
with gr.Row():
|
167 |
-
with gr.Column(scale=1):
|
168 |
-
input_model = gr.Dropdown(
|
169 |
-
MODEL_LIST, label="Model",
|
170 |
-
)
|
171 |
-
input_text = gr.TextArea(label="Text", info="The text to be extracted")
|
172 |
-
|
173 |
-
examples = gr.Examples(
|
174 |
-
examples=[
|
175 |
-
handle_text("""Legendary rock band Aerosmith has officially announced their retirement from touring after 54 years, citing
|
176 |
-
lead singer Steven Tyler's unrecoverable vocal cord injury.
|
177 |
-
The decision comes after months of unsuccessful treatment for Tyler's fractured larynx,
|
178 |
-
which he suffered in September 2023."""),
|
179 |
-
handle_text("""Pop star Justin Timberlake, 43, had his driver's license suspended by a New York judge during a virtual
|
180 |
-
court hearing on August 2, 2024. The suspension follows Timberlake's arrest for driving while intoxicated (DWI)
|
181 |
-
in Sag Harbor on June 18. Timberlake, who is currently on tour in Europe,
|
182 |
-
pleaded not guilty to the charges."""),
|
183 |
-
],
|
184 |
-
inputs=input_text
|
185 |
-
)
|
186 |
-
|
187 |
-
submit_button = gr.Button("Extract and Visualize")
|
188 |
|
189 |
-
|
190 |
-
|
191 |
-
|
192 |
-
|
193 |
-
|
194 |
-
|
195 |
-
|
196 |
-
|
197 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
198 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
199 |
|
|
|
200 |
demo.launch(share=False)
|
|
|
1 |
import spaces
|
2 |
import gradio as gr
|
3 |
from phi3_instruct_graph import MODEL_LIST, Phi3InstructGraph
|
|
|
4 |
import rapidjson
|
|
|
5 |
from pyvis.network import Network
|
6 |
import networkx as nx
|
7 |
import spacy
|
8 |
from spacy import displacy
|
9 |
from spacy.tokens import Span
|
10 |
import random
|
11 |
+
from tqdm import tqdm
|
12 |
+
|
13 |
+
# Constants
|
14 |
+
TITLE = "🌐 Phi-3 Instruct Graph Explorer"
|
15 |
+
SUBTITLE = "✨ Extract and visualize knowledge graphs from any text in multiple languages"
|
16 |
+
THEME = gr.themes.Base().set(
|
17 |
+
primary_hue="indigo",
|
18 |
+
secondary_hue="purple",
|
19 |
+
neutral_hue="slate",
|
20 |
+
radius_size=gr.themes.sizes.radius_sm,
|
21 |
+
shadow_size=gr.themes.sizes.shadow_lg,
|
22 |
+
)
|
23 |
+
|
24 |
+
# Color utilities
|
25 |
+
def get_random_light_color():
|
26 |
+
r = random.randint(140, 255)
|
27 |
+
g = random.randint(140, 255)
|
28 |
+
b = random.randint(140, 255)
|
29 |
+
return f"#{r:02x}{g:02x}{b:02x}"
|
30 |
|
31 |
+
# Text preprocessing
|
|
|
|
|
|
|
|
|
|
|
32 |
def handle_text(text):
|
33 |
return " ".join(text.split())
|
34 |
|
35 |
+
# Main processing functions
|
36 |
+
@spaces.GPU
|
37 |
+
def extract(text, model):
|
38 |
+
try:
|
39 |
+
model = Phi3InstructGraph(model=model)
|
40 |
+
result = model.extract(text)
|
41 |
+
return rapidjson.loads(result)
|
42 |
+
except Exception as e:
|
43 |
+
raise gr.Error(f"Extraction error: {str(e)}")
|
|
|
|
|
|
|
44 |
|
45 |
def find_token_indices(doc, substring, text):
|
46 |
result = []
|
|
|
57 |
if token.idx + len(token) == end_index:
|
58 |
end_token = token.i + 1
|
59 |
|
60 |
+
if start_token is not None and end_token is not None:
|
|
|
|
|
61 |
result.append({
|
62 |
"start": start_token,
|
63 |
"end": end_token
|
|
|
66 |
# Search for next occurrence
|
67 |
start_index = text.find(substring, end_index)
|
68 |
|
|
|
|
|
|
|
69 |
return result
|
70 |
|
|
|
71 |
def create_custom_entity_viz(data, full_text):
|
72 |
nlp = spacy.blank("xx")
|
73 |
doc = nlp(full_text)
|
|
|
85 |
overlapping = any(s.start < end and start < s.end for s in spans)
|
86 |
if not overlapping:
|
87 |
span = Span(doc, start, end, label=node["type"])
|
|
|
|
|
88 |
spans.append(span)
|
89 |
if node["type"] not in colors:
|
90 |
colors[node["type"]] = get_random_light_color()
|
|
|
102 |
html = displacy.render(doc, style="span", options=options)
|
103 |
return html
|
104 |
|
|
|
105 |
def create_graph(json_data):
|
106 |
G = nx.Graph()
|
107 |
|
108 |
+
# Add nodes with tooltips
|
109 |
for node in json_data['nodes']:
|
110 |
G.add_node(node['id'], title=f"{node['type']}: {node['detailed_type']}")
|
111 |
|
112 |
+
# Add edges with labels
|
113 |
for edge in json_data['edges']:
|
114 |
G.add_edge(edge['from'], edge['to'], title=edge['label'], label=edge['label'])
|
115 |
|
116 |
+
# Create network visualization
|
117 |
nt = Network(
|
118 |
width="720px",
|
119 |
height="600px",
|
120 |
directed=True,
|
121 |
notebook=False,
|
122 |
+
bgcolor="#f8fafc",
|
123 |
+
font_color="#1e293b"
|
|
|
|
|
124 |
)
|
125 |
+
|
126 |
+
# Configure network display
|
127 |
nt.from_nx(G)
|
128 |
nt.barnes_hut(
|
129 |
gravity=-3000,
|
|
|
133 |
damping=0.09,
|
134 |
overlap=0,
|
135 |
)
|
136 |
+
|
137 |
# Customize edge appearance
|
138 |
+
for edge in nt.edges:
|
139 |
+
edge['width'] = 2
|
140 |
+
edge['arrows'] = {'to': {'enabled': True, 'type': 'arrow'}}
|
141 |
+
edge['color'] = {'color': '#6366f1', 'highlight': '#4f46e5'}
|
142 |
+
edge['font'] = {'size': 12, 'color': '#4b5563', 'face': 'Arial'}
|
143 |
+
|
144 |
+
# Customize node appearance
|
145 |
+
for node in nt.nodes:
|
146 |
+
node['color'] = {'background': '#e0e7ff', 'border': '#6366f1', 'highlight': {'background': '#c7d2fe', 'border': '#4f46e5'}}
|
147 |
+
node['font'] = {'size': 14, 'color': '#1e293b'}
|
148 |
+
node['shape'] = 'dot'
|
149 |
+
node['size'] = 25
|
150 |
+
|
151 |
+
# Generate HTML with iframe to isolate styles
|
152 |
html = nt.generate_html()
|
153 |
html = html.replace("'", '"')
|
154 |
|
155 |
+
return f"""<iframe style="width: 100%; height: 620px; margin: 0 auto; border-radius: 8px; box-shadow: 0 4px 6px -1px rgba(0, 0, 0, 0.1), 0 2px 4px -1px rgba(0, 0, 0, 0.06);"
|
156 |
+
name="result" allow="midi; geolocation; microphone; camera; display-capture; encrypted-media;"
|
157 |
sandbox="allow-modals allow-forms allow-scripts allow-same-origin allow-popups
|
158 |
allow-top-navigation-by-user-activation allow-downloads" allowfullscreen=""
|
159 |
allowpaymentrequest="" frameborder="0" srcdoc='{html}'></iframe>"""
|
|
|
160 |
|
161 |
+
def process_and_visualize(text, model, progress=gr.Progress()):
|
162 |
if not text or not model:
|
163 |
+
raise gr.Error("⚠️ Both text and model must be provided.")
|
164 |
+
|
165 |
+
progress(0, desc="Starting extraction...")
|
166 |
json_data = extract(text, model)
|
167 |
+
|
168 |
+
progress(0.5, desc="Creating entity visualization...")
|
169 |
entities_viz = create_custom_entity_viz(json_data, text)
|
170 |
|
171 |
+
progress(0.8, desc="Building knowledge graph...")
|
172 |
graph_html = create_graph(json_data)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
173 |
|
174 |
+
node_count = len(json_data["nodes"])
|
175 |
+
edge_count = len(json_data["edges"])
|
176 |
+
stats = f"📊 Extracted {node_count} entities and {edge_count} relationships"
|
177 |
+
|
178 |
+
progress(1.0, desc="Complete!")
|
179 |
+
return graph_html, entities_viz, json_data, stats
|
180 |
+
|
181 |
+
# Example texts in different languages
|
182 |
+
EXAMPLES = [
|
183 |
+
[handle_text("""Legendary rock band Aerosmith has officially announced their retirement from touring after 54 years, citing
|
184 |
+
lead singer Steven Tyler's unrecoverable vocal cord injury.
|
185 |
+
The decision comes after months of unsuccessful treatment for Tyler's fractured larynx,
|
186 |
+
which he suffered in September 2023.""")],
|
187 |
+
|
188 |
+
[handle_text("""Pop star Justin Timberlake, 43, had his driver's license suspended by a New York judge during a virtual
|
189 |
+
court hearing on August 2, 2024. The suspension follows Timberlake's arrest for driving while intoxicated (DWI)
|
190 |
+
in Sag Harbor on June 18. Timberlake, who is currently on tour in Europe,
|
191 |
+
pleaded not guilty to the charges.""")],
|
192 |
|
193 |
+
[handle_text("""세계적인 기술 기업 삼성전자는 새로운 인공지능 기반 스마트폰을 올해 하반기에 출시할 예정이라고 발표했다.
|
194 |
+
이 스마트폰은 현재 개발 중인 갤럭시 시리즈의 최신작으로, 강력한 AI 기능과 혁신적인 카메라 시스템을 탑재할 것으로 알려졌다.
|
195 |
+
삼성전자의 CEO는 이번 신제품이 스마트폰 시장에 새로운 혁신을 가져올 것이라고 전망했다.""")],
|
196 |
+
|
197 |
+
[handle_text("""한국 영화 '기생충'은 2020년 아카데미 시상식에서 작품상, 감독상, 각본상, 국제영화상 등 4개 부문을 수상하며 역사를 새로 썼다.
|
198 |
+
봉준호 감독이 연출한 이 영화는 한국 영화 최초로 칸 영화제 황금종려상도 수상했으며, 전 세계적으로 엄청난 흥행과
|
199 |
+
평단의 호평을 받았다.""")]
|
200 |
+
]
|
201 |
+
|
202 |
+
def create_ui():
|
203 |
+
with gr.Blocks(theme=THEME, title=TITLE) as demo:
|
204 |
+
# Header
|
205 |
+
gr.Markdown(f"# {TITLE}")
|
206 |
+
gr.Markdown(f"{SUBTITLE}")
|
207 |
+
|
208 |
+
with gr.Row():
|
209 |
+
gr.Markdown("🌍 **Multilingual Support Available** 🔤")
|
210 |
+
|
211 |
+
# Main interface
|
212 |
+
with gr.Row():
|
213 |
+
# Input column
|
214 |
+
with gr.Column(scale=1):
|
215 |
+
input_model = gr.Dropdown(
|
216 |
+
MODEL_LIST,
|
217 |
+
label="🤖 Select Model",
|
218 |
+
info="Choose a model to process your text",
|
219 |
+
value=MODEL_LIST[0] if MODEL_LIST else None
|
220 |
+
)
|
221 |
+
|
222 |
+
input_text = gr.TextArea(
|
223 |
+
label="📝 Input Text",
|
224 |
+
info="Enter text in any language to extract a knowledge graph",
|
225 |
+
placeholder="Enter text here...",
|
226 |
+
lines=10
|
227 |
+
)
|
228 |
+
|
229 |
+
with gr.Row():
|
230 |
+
submit_button = gr.Button("🚀 Extract & Visualize", variant="primary", scale=2)
|
231 |
+
clear_button = gr.Button("🔄 Clear", variant="secondary", scale=1)
|
232 |
+
|
233 |
+
gr.Examples(
|
234 |
+
examples=EXAMPLES,
|
235 |
+
inputs=input_text,
|
236 |
+
label="📚 Example Texts (English & Korean)"
|
237 |
+
)
|
238 |
+
|
239 |
+
stats_output = gr.Markdown("", label="🔍 Analysis Results")
|
240 |
+
|
241 |
+
# Output column
|
242 |
+
with gr.Column(scale=1):
|
243 |
+
with gr.Tab("🧩 Knowledge Graph"):
|
244 |
+
output_graph = gr.HTML(label="")
|
245 |
+
|
246 |
+
with gr.Tab("🏷️ Entities"):
|
247 |
+
output_entity_viz = gr.HTML(label="")
|
248 |
+
|
249 |
+
with gr.Tab("📊 JSON Data"):
|
250 |
+
output_json = gr.JSON(label="")
|
251 |
+
|
252 |
+
# Functionality
|
253 |
+
submit_button.click(
|
254 |
+
fn=process_and_visualize,
|
255 |
+
inputs=[input_text, input_model],
|
256 |
+
outputs=[output_graph, output_entity_viz, output_json, stats_output]
|
257 |
+
)
|
258 |
+
|
259 |
+
clear_button.click(
|
260 |
+
fn=lambda: [None, None, None, ""],
|
261 |
+
inputs=[],
|
262 |
+
outputs=[output_graph, output_entity_viz, output_json, stats_output]
|
263 |
+
)
|
264 |
+
|
265 |
+
# Footer
|
266 |
+
gr.Markdown("---")
|
267 |
+
gr.Markdown("📋 **Instructions:** Enter text in any language, select a model, and click 'Extract & Visualize' to generate a knowledge graph.")
|
268 |
+
gr.Markdown("🛠️ Powered by Phi-3 Instruct Graph | Created by Emergent Methods")
|
269 |
+
|
270 |
+
return demo
|
271 |
|
272 |
+
demo = create_ui()
|
273 |
demo.launch(share=False)
|