Create examples.py
Browse files- examples.py +731 -0
examples.py
ADDED
@@ -0,0 +1,731 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
"""
|
2 |
+
π Perplexity AI Integration Examples
|
3 |
+
Demonstrate how to effectively use AI-powered source discovery for dataset creation
|
4 |
+
"""
|
5 |
+
|
6 |
+
import os
|
7 |
+
import json
|
8 |
+
import time
|
9 |
+
from typing import List, Dict
|
10 |
+
from datetime import datetime
|
11 |
+
|
12 |
+
# Import our Perplexity client
|
13 |
+
try:
|
14 |
+
from perplexity_client import PerplexityClient, SearchType, SourceResult
|
15 |
+
PERPLEXITY_AVAILABLE = True
|
16 |
+
except ImportError:
|
17 |
+
print("β οΈ Perplexity client not available. Make sure perplexity_client.py is in the same directory.")
|
18 |
+
PERPLEXITY_AVAILABLE = False
|
19 |
+
|
20 |
+
def example_sentiment_analysis_sources():
|
21 |
+
"""
|
22 |
+
π Example: Find sources for sentiment analysis dataset
|
23 |
+
|
24 |
+
This example shows how to discover diverse sources for sentiment analysis,
|
25 |
+
including product reviews, social media, and news content.
|
26 |
+
"""
|
27 |
+
print("π Example: Sentiment Analysis Source Discovery")
|
28 |
+
print("=" * 60)
|
29 |
+
|
30 |
+
if not PERPLEXITY_AVAILABLE:
|
31 |
+
print("β Perplexity client not available")
|
32 |
+
return
|
33 |
+
|
34 |
+
client = PerplexityClient()
|
35 |
+
|
36 |
+
if not client._validate_api_key():
|
37 |
+
print("β Please set PERPLEXITY_API_KEY environment variable")
|
38 |
+
return
|
39 |
+
|
40 |
+
# Different types of sentiment analysis projects
|
41 |
+
projects = [
|
42 |
+
{
|
43 |
+
"description": "Product reviews from e-commerce sites for sentiment classification of customer feedback",
|
44 |
+
"search_type": SearchType.GENERAL,
|
45 |
+
"focus": "E-commerce reviews"
|
46 |
+
},
|
47 |
+
{
|
48 |
+
"description": "Movie and entertainment reviews for sentiment analysis training with detailed ratings",
|
49 |
+
"search_type": SearchType.GENERAL,
|
50 |
+
"focus": "Entertainment reviews"
|
51 |
+
},
|
52 |
+
{
|
53 |
+
"description": "Social media posts and comments about brands for real-time sentiment monitoring",
|
54 |
+
"search_type": SearchType.SOCIAL,
|
55 |
+
"focus": "Social media sentiment"
|
56 |
+
},
|
57 |
+
{
|
58 |
+
"description": "News articles with opinion content for political sentiment analysis research",
|
59 |
+
"search_type": SearchType.NEWS,
|
60 |
+
"focus": "News opinion analysis"
|
61 |
+
}
|
62 |
+
]
|
63 |
+
|
64 |
+
all_results = []
|
65 |
+
|
66 |
+
for i, project in enumerate(projects, 1):
|
67 |
+
print(f"\nπ Project {i}: {project['focus']}")
|
68 |
+
print("-" * 40)
|
69 |
+
|
70 |
+
try:
|
71 |
+
results = client.discover_sources(
|
72 |
+
project_description=project["description"],
|
73 |
+
search_type=project["search_type"],
|
74 |
+
max_sources=8,
|
75 |
+
include_academic=False, # Focus on practical sources
|
76 |
+
include_news=True
|
77 |
+
)
|
78 |
+
|
79 |
+
print(f"β
Found {len(results.sources)} sources in {results.search_time:.1f}s")
|
80 |
+
|
81 |
+
# Show top 3 sources
|
82 |
+
for j, source in enumerate(results.sources[:3], 1):
|
83 |
+
print(f" {j}. {source.title}")
|
84 |
+
print(f" URL: {source.url}")
|
85 |
+
print(f" Type: {source.source_type} | Score: {source.relevance_score:.1f}/10")
|
86 |
+
print(f" Description: {source.description[:100]}...")
|
87 |
+
print()
|
88 |
+
|
89 |
+
all_results.extend(results.sources)
|
90 |
+
|
91 |
+
if results.suggestions:
|
92 |
+
print(f"π‘ Suggestions: {', '.join(results.suggestions[:3])}")
|
93 |
+
|
94 |
+
except Exception as e:
|
95 |
+
print(f"β Error: {e}")
|
96 |
+
|
97 |
+
# Respectful delay between requests
|
98 |
+
time.sleep(1)
|
99 |
+
|
100 |
+
# Summary
|
101 |
+
print(f"\nπ SUMMARY")
|
102 |
+
print("-" * 40)
|
103 |
+
print(f"Total sources discovered: {len(all_results)}")
|
104 |
+
|
105 |
+
# Analyze source types
|
106 |
+
source_types = {}
|
107 |
+
for source in all_results:
|
108 |
+
source_types[source.source_type] = source_types.get(source.source_type, 0) + 1
|
109 |
+
|
110 |
+
print("Source type distribution:")
|
111 |
+
for stype, count in sorted(source_types.items()):
|
112 |
+
print(f" {stype}: {count} sources")
|
113 |
+
|
114 |
+
# Top domains
|
115 |
+
domains = {}
|
116 |
+
for source in all_results:
|
117 |
+
domains[source.domain] = domains.get(source.domain, 0) + 1
|
118 |
+
|
119 |
+
print("\nTop domains:")
|
120 |
+
for domain, count in sorted(domains.items(), key=lambda x: x[1], reverse=True)[:5]:
|
121 |
+
print(f" {domain}: {count} sources")
|
122 |
+
|
123 |
+
return all_results
|
124 |
+
|
125 |
+
def example_text_classification_sources():
|
126 |
+
"""
|
127 |
+
π Example: Find sources for text classification dataset
|
128 |
+
|
129 |
+
This example demonstrates finding well-categorized content for
|
130 |
+
multi-class text classification training.
|
131 |
+
"""
|
132 |
+
print("\nπ Example: Text Classification Source Discovery")
|
133 |
+
print("=" * 60)
|
134 |
+
|
135 |
+
if not PERPLEXITY_AVAILABLE:
|
136 |
+
print("β Perplexity client not available")
|
137 |
+
return
|
138 |
+
|
139 |
+
client = PerplexityClient()
|
140 |
+
|
141 |
+
# Multi-domain classification project
|
142 |
+
project_description = """
|
143 |
+
Find diverse news articles and content with clear topical categories for training
|
144 |
+
a multi-class text classifier. Need sources covering politics, technology, sports,
|
145 |
+
business, entertainment, health, and science topics with consistent categorization.
|
146 |
+
"""
|
147 |
+
|
148 |
+
try:
|
149 |
+
results = client.discover_sources(
|
150 |
+
project_description=project_description,
|
151 |
+
search_type=SearchType.NEWS,
|
152 |
+
max_sources=15,
|
153 |
+
include_academic=True, # Include academic sources for science topics
|
154 |
+
include_news=True
|
155 |
+
)
|
156 |
+
|
157 |
+
print(f"β
Found {len(results.sources)} sources")
|
158 |
+
|
159 |
+
# Categorize sources by likely content type
|
160 |
+
categorized = {
|
161 |
+
"news": [],
|
162 |
+
"academic": [],
|
163 |
+
"business": [],
|
164 |
+
"technology": [],
|
165 |
+
"other": []
|
166 |
+
}
|
167 |
+
|
168 |
+
for source in results.sources:
|
169 |
+
domain = source.domain.lower()
|
170 |
+
if any(news in domain for news in ['reuters', 'bbc', 'cnn', 'news']):
|
171 |
+
categorized["news"].append(source)
|
172 |
+
elif any(academic in domain for academic in ['arxiv', 'pubmed', 'scholar', 'edu']):
|
173 |
+
categorized["academic"].append(source)
|
174 |
+
elif any(biz in domain for biz in ['bloomberg', 'forbes', 'business', 'financial']):
|
175 |
+
categorized["business"].append(source)
|
176 |
+
elif any(tech in domain for tech in ['techcrunch', 'wired', 'tech', 'digital']):
|
177 |
+
categorized["technology"].append(source)
|
178 |
+
else:
|
179 |
+
categorized["other"].append(source)
|
180 |
+
|
181 |
+
print("\nπ Sources by Category:")
|
182 |
+
for category, sources in categorized.items():
|
183 |
+
if sources:
|
184 |
+
print(f"\n{category.upper()} ({len(sources)} sources):")
|
185 |
+
for source in sources[:2]: # Show top 2 per category
|
186 |
+
print(f" β’ {source.title}")
|
187 |
+
print(f" {source.url}")
|
188 |
+
print(f" Score: {source.relevance_score:.1f}/10")
|
189 |
+
|
190 |
+
# Export for use
|
191 |
+
export_data = client.export_sources(results, "json")
|
192 |
+
|
193 |
+
# Save to file
|
194 |
+
filename = f"text_classification_sources_{int(time.time())}.json"
|
195 |
+
with open(filename, 'w', encoding='utf-8') as f:
|
196 |
+
f.write(export_data)
|
197 |
+
|
198 |
+
print(f"\nπ Sources exported to: {filename}")
|
199 |
+
|
200 |
+
return results.sources
|
201 |
+
|
202 |
+
except Exception as e:
|
203 |
+
print(f"β Error: {e}")
|
204 |
+
return []
|
205 |
+
|
206 |
+
def example_academic_research_sources():
|
207 |
+
"""
|
208 |
+
π Example: Find academic sources for research dataset
|
209 |
+
|
210 |
+
This example shows how to discover high-quality academic sources
|
211 |
+
for research-focused datasets.
|
212 |
+
"""
|
213 |
+
print("\nπ Example: Academic Research Source Discovery")
|
214 |
+
print("=" * 60)
|
215 |
+
|
216 |
+
if not PERPLEXITY_AVAILABLE:
|
217 |
+
print("β Perplexity client not available")
|
218 |
+
return
|
219 |
+
|
220 |
+
client = PerplexityClient()
|
221 |
+
|
222 |
+
# Research-focused projects
|
223 |
+
research_topics = [
|
224 |
+
{
|
225 |
+
"description": "Recent machine learning research papers on transformer architectures and attention mechanisms for NLP survey dataset",
|
226 |
+
"domain_focus": "AI/ML research"
|
227 |
+
},
|
228 |
+
{
|
229 |
+
"description": "Climate change research papers and reports for environmental science text summarization training",
|
230 |
+
"domain_focus": "Climate science"
|
231 |
+
},
|
232 |
+
{
|
233 |
+
"description": "Medical research papers on drug discovery and pharmaceutical research for biomedical NER training",
|
234 |
+
"domain_focus": "Medical research"
|
235 |
+
}
|
236 |
+
]
|
237 |
+
|
238 |
+
all_academic_sources = []
|
239 |
+
|
240 |
+
for topic in research_topics:
|
241 |
+
print(f"\n㪠Research Topic: {topic['domain_focus']}")
|
242 |
+
print("-" * 40)
|
243 |
+
|
244 |
+
try:
|
245 |
+
results = client.discover_sources(
|
246 |
+
project_description=topic["description"],
|
247 |
+
search_type=SearchType.ACADEMIC,
|
248 |
+
max_sources=10,
|
249 |
+
include_academic=True,
|
250 |
+
include_news=False # Focus on academic sources only
|
251 |
+
)
|
252 |
+
|
253 |
+
print(f"β
Found {len(results.sources)} academic sources")
|
254 |
+
|
255 |
+
# Filter for high-quality academic sources
|
256 |
+
high_quality = [s for s in results.sources if s.relevance_score >= 7.0]
|
257 |
+
|
258 |
+
print(f"π High-quality sources (score β₯ 7.0): {len(high_quality)}")
|
259 |
+
|
260 |
+
for source in high_quality[:3]:
|
261 |
+
print(f"\n π {source.title}")
|
262 |
+
print(f" URL: {source.url}")
|
263 |
+
print(f" Domain: {source.domain}")
|
264 |
+
print(f" Score: {source.relevance_score:.1f}/10")
|
265 |
+
print(f" Type: {source.source_type}")
|
266 |
+
|
267 |
+
all_academic_sources.extend(high_quality)
|
268 |
+
|
269 |
+
except Exception as e:
|
270 |
+
print(f"β Error: {e}")
|
271 |
+
|
272 |
+
time.sleep(1) # Respectful delay
|
273 |
+
|
274 |
+
# Analysis
|
275 |
+
print(f"\nπ ACADEMIC SOURCES ANALYSIS")
|
276 |
+
print("-" * 40)
|
277 |
+
print(f"Total high-quality academic sources: {len(all_academic_sources)}")
|
278 |
+
|
279 |
+
# Domain analysis
|
280 |
+
academic_domains = {}
|
281 |
+
for source in all_academic_sources:
|
282 |
+
domain = source.domain
|
283 |
+
academic_domains[domain] = academic_domains.get(domain, 0) + 1
|
284 |
+
|
285 |
+
print("\nTop academic domains:")
|
286 |
+
for domain, count in sorted(academic_domains.items(), key=lambda x: x[1], reverse=True)[:5]:
|
287 |
+
print(f" {domain}: {count} papers")
|
288 |
+
|
289 |
+
# Quality distribution
|
290 |
+
scores = [s.relevance_score for s in all_academic_sources]
|
291 |
+
if scores:
|
292 |
+
avg_score = sum(scores) / len(scores)
|
293 |
+
print(f"\nAverage quality score: {avg_score:.1f}/10")
|
294 |
+
print(f"Score range: {min(scores):.1f} - {max(scores):.1f}")
|
295 |
+
|
296 |
+
return all_academic_sources
|
297 |
+
|
298 |
+
def example_custom_search_strategies():
|
299 |
+
"""
|
300 |
+
π― Example: Custom search strategies for specific needs
|
301 |
+
|
302 |
+
This example demonstrates advanced techniques for finding
|
303 |
+
very specific types of content.
|
304 |
+
"""
|
305 |
+
print("\nπ― Example: Custom Search Strategies")
|
306 |
+
print("=" * 60)
|
307 |
+
|
308 |
+
if not PERPLEXITY_AVAILABLE:
|
309 |
+
print("β Perplexity client not available")
|
310 |
+
return
|
311 |
+
|
312 |
+
client = PerplexityClient()
|
313 |
+
|
314 |
+
# Strategy 1: Domain-specific search
|
315 |
+
print("\nπ Strategy 1: Domain-specific Financial Content")
|
316 |
+
print("-" * 50)
|
317 |
+
|
318 |
+
try:
|
319 |
+
financial_results = client.get_domain_sources(
|
320 |
+
domain="bloomberg.com",
|
321 |
+
topic="quarterly earnings reports and financial analysis",
|
322 |
+
max_sources=5
|
323 |
+
)
|
324 |
+
|
325 |
+
print(f"β
Found {len(financial_results.sources)} financial sources")
|
326 |
+
for source in financial_results.sources[:2]:
|
327 |
+
print(f" β’ {source.title}")
|
328 |
+
print(f" Score: {source.relevance_score:.1f}/10")
|
329 |
+
|
330 |
+
except Exception as e:
|
331 |
+
print(f"β Error: {e}")
|
332 |
+
|
333 |
+
# Strategy 2: Keyword-based search
|
334 |
+
print("\nπ Strategy 2: Keyword-based Technical Content")
|
335 |
+
print("-" * 50)
|
336 |
+
|
337 |
+
try:
|
338 |
+
tech_keywords = ["API documentation", "software tutorials", "programming guides", "technical specifications"]
|
339 |
+
tech_results = client.search_with_keywords(
|
340 |
+
keywords=tech_keywords,
|
341 |
+
search_type=SearchType.TECHNICAL
|
342 |
+
)
|
343 |
+
|
344 |
+
print(f"β
Found {len(tech_results.sources)} technical sources")
|
345 |
+
for source in tech_results.sources[:2]:
|
346 |
+
print(f" β’ {source.title}")
|
347 |
+
print(f" Type: {source.source_type}")
|
348 |
+
|
349 |
+
except Exception as e:
|
350 |
+
print(f"β Error: {e}")
|
351 |
+
|
352 |
+
# Strategy 3: Multi-format search
|
353 |
+
print("\nπ Strategy 3: Multi-format Content Discovery")
|
354 |
+
print("-" * 50)
|
355 |
+
|
356 |
+
multiformat_description = """
|
357 |
+
Find diverse content formats including FAQ pages, interview transcripts,
|
358 |
+
tutorial content, and documentation for question-answering dataset creation.
|
359 |
+
Need sources with clear question-answer patterns and structured information.
|
360 |
+
"""
|
361 |
+
|
362 |
+
try:
|
363 |
+
qa_results = client.discover_sources(
|
364 |
+
project_description=multiformat_description,
|
365 |
+
search_type=SearchType.GENERAL,
|
366 |
+
max_sources=12
|
367 |
+
)
|
368 |
+
|
369 |
+
print(f"β
Found {len(qa_results.sources)} Q&A sources")
|
370 |
+
|
371 |
+
# Categorize by content format
|
372 |
+
formats = {
|
373 |
+
"faq": [],
|
374 |
+
"tutorial": [],
|
375 |
+
"documentation": [],
|
376 |
+
"interview": [],
|
377 |
+
"other": []
|
378 |
+
}
|
379 |
+
|
380 |
+
for source in qa_results.sources:
|
381 |
+
title_lower = source.title.lower()
|
382 |
+
url_lower = source.url.lower()
|
383 |
+
|
384 |
+
if any(faq in title_lower or faq in url_lower for faq in ['faq', 'questions', 'help']):
|
385 |
+
formats["faq"].append(source)
|
386 |
+
elif any(tut in title_lower for tut in ['tutorial', 'guide', 'how to']):
|
387 |
+
formats["tutorial"].append(source)
|
388 |
+
elif any(doc in title_lower or doc in url_lower for doc in ['docs', 'documentation', 'manual']):
|
389 |
+
formats["documentation"].append(source)
|
390 |
+
elif any(int in title_lower for int in ['interview', 'q&a', 'conversation']):
|
391 |
+
formats["interview"].append(source)
|
392 |
+
else:
|
393 |
+
formats["other"].append(source)
|
394 |
+
|
395 |
+
for format_type, sources in formats.items():
|
396 |
+
if sources:
|
397 |
+
print(f"\n {format_type.upper()}: {len(sources)} sources")
|
398 |
+
if sources:
|
399 |
+
best = max(sources, key=lambda x: x.relevance_score)
|
400 |
+
print(f" Best: {best.title} (Score: {best.relevance_score:.1f})")
|
401 |
+
|
402 |
+
except Exception as e:
|
403 |
+
print(f"β Error: {e}")
|
404 |
+
|
405 |
+
def example_quality_assessment():
|
406 |
+
"""
|
407 |
+
β
Example: Quality assessment and source validation
|
408 |
+
|
409 |
+
This example shows how to evaluate and filter sources
|
410 |
+
for maximum dataset quality.
|
411 |
+
"""
|
412 |
+
print("\nβ
Example: Source Quality Assessment")
|
413 |
+
print("=" * 60)
|
414 |
+
|
415 |
+
if not PERPLEXITY_AVAILABLE:
|
416 |
+
print("β Perplexity client not available")
|
417 |
+
return
|
418 |
+
|
419 |
+
client = PerplexityClient()
|
420 |
+
|
421 |
+
# Broad search to get diverse quality levels
|
422 |
+
description = "Content for machine learning training including text classification and sentiment analysis"
|
423 |
+
|
424 |
+
try:
|
425 |
+
results = client.discover_sources(
|
426 |
+
project_description=description,
|
427 |
+
search_type=SearchType.GENERAL,
|
428 |
+
max_sources=20
|
429 |
+
)
|
430 |
+
|
431 |
+
print(f"β
Found {len(results.sources)} total sources")
|
432 |
+
|
433 |
+
# Quality analysis
|
434 |
+
print(f"\nπ QUALITY DISTRIBUTION")
|
435 |
+
print("-" * 40)
|
436 |
+
|
437 |
+
quality_tiers = {
|
438 |
+
"excellent": [s for s in results.sources if s.relevance_score >= 8.0],
|
439 |
+
"good": [s for s in results.sources if 6.0 <= s.relevance_score < 8.0],
|
440 |
+
"acceptable": [s for s in results.sources if 4.0 <= s.relevance_score < 6.0],
|
441 |
+
"poor": [s for s in results.sources if s.relevance_score < 4.0]
|
442 |
+
}
|
443 |
+
|
444 |
+
for tier, sources in quality_tiers.items():
|
445 |
+
print(f"{tier.upper()}: {len(sources)} sources")
|
446 |
+
if sources:
|
447 |
+
avg_score = sum(s.relevance_score for s in sources) / len(sources)
|
448 |
+
print(f" Average score: {avg_score:.1f}")
|
449 |
+
print(f" Example: {sources[0].title[:50]}...")
|
450 |
+
|
451 |
+
# Validate top sources
|
452 |
+
print(f"\nπ VALIDATING TOP SOURCES")
|
453 |
+
print("-" * 40)
|
454 |
+
|
455 |
+
top_sources = [s for s in results.sources if s.relevance_score >= 7.0]
|
456 |
+
validated_sources = client.validate_sources(top_sources)
|
457 |
+
|
458 |
+
print(f"Sources passed validation: {len(validated_sources)}/{len(top_sources)}")
|
459 |
+
|
460 |
+
# Show validation results
|
461 |
+
for source in validated_sources[:3]:
|
462 |
+
print(f"\nβ
VALIDATED: {source.title}")
|
463 |
+
print(f" URL: {source.url}")
|
464 |
+
print(f" Domain: {source.domain}")
|
465 |
+
print(f" Type: {source.source_type}")
|
466 |
+
print(f" Score: {source.relevance_score:.1f}/10")
|
467 |
+
print(f" Description: {source.description[:100]}...")
|
468 |
+
|
469 |
+
# Export validated sources
|
470 |
+
if validated_sources:
|
471 |
+
export_data = {
|
472 |
+
"search_query": description,
|
473 |
+
"total_found": len(results.sources),
|
474 |
+
"validated_count": len(validated_sources),
|
475 |
+
"quality_threshold": 7.0,
|
476 |
+
"sources": [
|
477 |
+
{
|
478 |
+
"url": s.url,
|
479 |
+
"title": s.title,
|
480 |
+
"domain": s.domain,
|
481 |
+
"type": s.source_type,
|
482 |
+
"score": s.relevance_score,
|
483 |
+
"description": s.description
|
484 |
+
}
|
485 |
+
for s in validated_sources
|
486 |
+
]
|
487 |
+
}
|
488 |
+
|
489 |
+
filename = f"validated_sources_{int(time.time())}.json"
|
490 |
+
with open(filename, 'w', encoding='utf-8') as f:
|
491 |
+
json.dump(export_data, f, indent=2)
|
492 |
+
|
493 |
+
print(f"\nπ Validated sources exported to: {filename}")
|
494 |
+
|
495 |
+
return validated_sources
|
496 |
+
|
497 |
+
except Exception as e:
|
498 |
+
print(f"β Error: {e}")
|
499 |
+
return []
|
500 |
+
|
501 |
+
def example_batch_processing():
|
502 |
+
"""
|
503 |
+
β‘ Example: Batch processing for large dataset projects
|
504 |
+
|
505 |
+
This example demonstrates efficient batch discovery for
|
506 |
+
large-scale dataset creation projects.
|
507 |
+
"""
|
508 |
+
print("\nβ‘ Example: Batch Processing for Large Projects")
|
509 |
+
print("=" * 60)
|
510 |
+
|
511 |
+
if not PERPLEXITY_AVAILABLE:
|
512 |
+
print("β Perplexity client not available")
|
513 |
+
return
|
514 |
+
|
515 |
+
client = PerplexityClient()
|
516 |
+
|
517 |
+
# Define multiple related searches for comprehensive coverage
|
518 |
+
batch_searches = [
|
519 |
+
{
|
520 |
+
"name": "E-commerce Reviews",
|
521 |
+
"description": "Product reviews from online stores for sentiment analysis",
|
522 |
+
"search_type": SearchType.GENERAL,
|
523 |
+
"max_sources": 8
|
524 |
+
},
|
525 |
+
{
|
526 |
+
"name": "Social Media Content",
|
527 |
+
"description": "Social media posts and comments for sentiment classification",
|
528 |
+
"search_type": SearchType.SOCIAL,
|
529 |
+
"max_sources": 8
|
530 |
+
},
|
531 |
+
{
|
532 |
+
"name": "News Opinion",
|
533 |
+
"description": "News articles with editorial content for opinion mining",
|
534 |
+
"search_type": SearchType.NEWS,
|
535 |
+
"max_sources": 8
|
536 |
+
},
|
537 |
+
{
|
538 |
+
"name": "Forum Discussions",
|
539 |
+
"description": "Forum posts and community discussions for sentiment analysis",
|
540 |
+
"search_type": SearchType.GENERAL,
|
541 |
+
"max_sources": 6
|
542 |
+
}
|
543 |
+
]
|
544 |
+
|
545 |
+
all_batch_results = []
|
546 |
+
total_start_time = time.time()
|
547 |
+
|
548 |
+
print(f"π Processing {len(batch_searches)} batch searches...")
|
549 |
+
|
550 |
+
for i, search in enumerate(batch_searches, 1):
|
551 |
+
print(f"\nπ Batch {i}/{len(batch_searches)}: {search['name']}")
|
552 |
+
print("-" * 40)
|
553 |
+
|
554 |
+
search_start = time.time()
|
555 |
+
|
556 |
+
try:
|
557 |
+
results = client.discover_sources(
|
558 |
+
project_description=search["description"],
|
559 |
+
search_type=search["search_type"],
|
560 |
+
max_sources=search["max_sources"]
|
561 |
+
)
|
562 |
+
|
563 |
+
search_time = time.time() - search_start
|
564 |
+
|
565 |
+
print(f"β
Found {len(results.sources)} sources in {search_time:.1f}s")
|
566 |
+
|
567 |
+
# Add batch metadata
|
568 |
+
for source in results.sources:
|
569 |
+
source.batch_name = search["name"]
|
570 |
+
source.batch_index = i
|
571 |
+
|
572 |
+
all_batch_results.extend(results.sources)
|
573 |
+
|
574 |
+
# Show top result
|
575 |
+
if results.sources:
|
576 |
+
best = max(results.sources, key=lambda x: x.relevance_score)
|
577 |
+
print(f" Top result: {best.title} (Score: {best.relevance_score:.1f})")
|
578 |
+
|
579 |
+
except Exception as e:
|
580 |
+
print(f"β Batch {i} failed: {e}")
|
581 |
+
|
582 |
+
# Rate limiting between batches
|
583 |
+
time.sleep(1.5)
|
584 |
+
|
585 |
+
total_time = time.time() - total_start_time
|
586 |
+
|
587 |
+
# Batch results analysis
|
588 |
+
print(f"\nπ BATCH PROCESSING RESULTS")
|
589 |
+
print("-" * 40)
|
590 |
+
print(f"Total sources discovered: {len(all_batch_results)}")
|
591 |
+
print(f"Total processing time: {total_time:.1f} seconds")
|
592 |
+
print(f"Average per batch: {total_time/len(batch_searches):.1f} seconds")
|
593 |
+
|
594 |
+
# Quality distribution across batches
|
595 |
+
batch_stats = {}
|
596 |
+
for source in all_batch_results:
|
597 |
+
batch_name = getattr(source, 'batch_name', 'unknown')
|
598 |
+
if batch_name not in batch_stats:
|
599 |
+
batch_stats[batch_name] = {
|
600 |
+
'count': 0,
|
601 |
+
'avg_score': 0,
|
602 |
+
'scores': []
|
603 |
+
}
|
604 |
+
|
605 |
+
batch_stats[batch_name]['count'] += 1
|
606 |
+
batch_stats[batch_name]['scores'].append(source.relevance_score)
|
607 |
+
|
608 |
+
# Calculate averages
|
609 |
+
for batch_name, stats in batch_stats.items():
|
610 |
+
if stats['scores']:
|
611 |
+
stats['avg_score'] = sum(stats['scores']) / len(stats['scores'])
|
612 |
+
|
613 |
+
print(f"\nBatch quality comparison:")
|
614 |
+
for batch_name, stats in sorted(batch_stats.items(), key=lambda x: x[1]['avg_score'], reverse=True):
|
615 |
+
print(f" {batch_name}: {stats['count']} sources, avg score {stats['avg_score']:.1f}")
|
616 |
+
|
617 |
+
# Export comprehensive results
|
618 |
+
batch_export = {
|
619 |
+
"project_name": "Large Scale Sentiment Analysis Dataset",
|
620 |
+
"batch_processing_date": datetime.now().isoformat(),
|
621 |
+
"total_sources": len(all_batch_results),
|
622 |
+
"processing_time_seconds": total_time,
|
623 |
+
"batches": len(batch_searches),
|
624 |
+
"batch_statistics": batch_stats,
|
625 |
+
"sources": [
|
626 |
+
{
|
627 |
+
"url": s.url,
|
628 |
+
"title": s.title,
|
629 |
+
"domain": s.domain,
|
630 |
+
"type": s.source_type,
|
631 |
+
"score": s.relevance_score,
|
632 |
+
"batch": getattr(s, 'batch_name', 'unknown'),
|
633 |
+
"description": s.description
|
634 |
+
}
|
635 |
+
for s in all_batch_results
|
636 |
+
]
|
637 |
+
}
|
638 |
+
|
639 |
+
filename = f"batch_results_{int(time.time())}.json"
|
640 |
+
with open(filename, 'w', encoding='utf-8') as f:
|
641 |
+
json.dump(batch_export, f, indent=2)
|
642 |
+
|
643 |
+
print(f"\nπ Batch results exported to: {filename}")
|
644 |
+
print(f"π‘ Use these {len(all_batch_results)} sources to create a comprehensive sentiment analysis dataset!")
|
645 |
+
|
646 |
+
return all_batch_results
|
647 |
+
|
648 |
+
def main():
|
649 |
+
"""
|
650 |
+
π Run all Perplexity AI examples
|
651 |
+
|
652 |
+
This function demonstrates the full range of capabilities
|
653 |
+
for AI-powered source discovery.
|
654 |
+
"""
|
655 |
+
print("π Perplexity AI Integration - Complete Examples")
|
656 |
+
print("=" * 70)
|
657 |
+
print("These examples show how to use AI-powered source discovery")
|
658 |
+
print("to create high-quality datasets efficiently.\n")
|
659 |
+
|
660 |
+
if not PERPLEXITY_AVAILABLE:
|
661 |
+
print("β Cannot run examples - perplexity_client.py not found")
|
662 |
+
print("Please ensure the perplexity_client.py file is in the same directory.")
|
663 |
+
return
|
664 |
+
|
665 |
+
if not os.getenv('PERPLEXITY_API_KEY'):
|
666 |
+
print("β Cannot run examples - PERPLEXITY_API_KEY not set")
|
667 |
+
print("Please set your Perplexity API key as an environment variable:")
|
668 |
+
print("export PERPLEXITY_API_KEY='your_api_key_here'")
|
669 |
+
return
|
670 |
+
|
671 |
+
print("β
Perplexity AI client available and configured")
|
672 |
+
print("π― Running comprehensive examples...\n")
|
673 |
+
|
674 |
+
try:
|
675 |
+
# Run all examples
|
676 |
+
sentiment_sources = example_sentiment_analysis_sources()
|
677 |
+
time.sleep(2) # Respectful delay
|
678 |
+
|
679 |
+
classification_sources = example_text_classification_sources()
|
680 |
+
time.sleep(2)
|
681 |
+
|
682 |
+
academic_sources = example_academic_research_sources()
|
683 |
+
time.sleep(2)
|
684 |
+
|
685 |
+
example_custom_search_strategies()
|
686 |
+
time.sleep(2)
|
687 |
+
|
688 |
+
validated_sources = example_quality_assessment()
|
689 |
+
time.sleep(2)
|
690 |
+
|
691 |
+
batch_sources = example_batch_processing()
|
692 |
+
|
693 |
+
# Final summary
|
694 |
+
print(f"\nπ EXAMPLES COMPLETE!")
|
695 |
+
print("=" * 70)
|
696 |
+
print("Summary of discovered sources:")
|
697 |
+
|
698 |
+
total_sources = 0
|
699 |
+
if sentiment_sources:
|
700 |
+
total_sources += len(sentiment_sources)
|
701 |
+
print(f" π Sentiment Analysis: {len(sentiment_sources)} sources")
|
702 |
+
|
703 |
+
if classification_sources:
|
704 |
+
total_sources += len(classification_sources)
|
705 |
+
print(f" π Text Classification: {len(classification_sources)} sources")
|
706 |
+
|
707 |
+
if academic_sources:
|
708 |
+
total_sources += len(academic_sources)
|
709 |
+
print(f" π Academic Research: {len(academic_sources)} sources")
|
710 |
+
|
711 |
+
if validated_sources:
|
712 |
+
print(f" β
Validated High-Quality: {len(validated_sources)} sources")
|
713 |
+
|
714 |
+
if batch_sources:
|
715 |
+
print(f" β‘ Batch Processing: {len(batch_sources)} sources")
|
716 |
+
|
717 |
+
print(f"\nπ― Total unique sources discovered: {total_sources}")
|
718 |
+
print("π Check the generated JSON files for detailed source information")
|
719 |
+
print("\nπ‘ Next steps:")
|
720 |
+
print(" 1. Review the exported source files")
|
721 |
+
print(" 2. Select the best sources for your specific use case")
|
722 |
+
print(" 3. Use these sources in your AI Dataset Studio")
|
723 |
+
print(" 4. Create amazing datasets with AI-powered discovery!")
|
724 |
+
|
725 |
+
except Exception as e:
|
726 |
+
print(f"β Error running examples: {e}")
|
727 |
+
import traceback
|
728 |
+
traceback.print_exc()
|
729 |
+
|
730 |
+
if __name__ == "__main__":
|
731 |
+
main()
|