Spaces:
Sleeping
Sleeping
File size: 32,029 Bytes
7bf325d 8e34de3 b0473cc 6528c77 fe68698 33f6fed fe68698 6aa7fe7 fe68698 8738bd2 fe68698 8738bd2 fe68698 36f7a03 6b7b8b5 36f7a03 8e34de3 36f7a03 8e34de3 9a806ac c435293 9a806ac c435293 9a806ac c435293 9a806ac 08f222a 9a806ac 08f222a 9a806ac 08f222a 9a806ac 08f222a 9a806ac 08f222a 9a806ac 08f222a 9a806ac 08f222a 9a806ac 08f222a 9a806ac 08f222a 9a806ac 08f222a 9a806ac c435293 9a806ac c435293 9a806ac a800293 9a806ac a800293 9a806ac c435293 9a806ac c435293 9a806ac c435293 9a806ac c435293 9a806ac c435293 9a806ac c435293 9a806ac c435293 9a806ac a800293 9a806ac c435293 9a806ac c435293 9a806ac a800293 9a806ac c435293 9a806ac a800293 9a806ac 6aa7fe7 9a806ac ce0a41c 9a806ac 6aa7fe7 9a806ac c435293 6aa7fe7 c435293 6aa7fe7 c435293 6aa7fe7 c435293 6aa7fe7 c435293 6aa7fe7 c435293 ab35b41 cc57712 36f7a03 8e34de3 36f7a03 fe68698 8e34de3 a800293 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 |
import gradio as gr
from ui.dataset_input import create_dataset_input, load_example_dataset
from ui.analysis_screen import create_analysis_screen, process_analysis_request
from visualization.bow_visualizer import process_and_visualize_analysis
import nltk
import os
import json
# Download necessary NLTK resources function remains unchanged
def download_nltk_resources():
"""Download required NLTK resources if not already downloaded"""
try:
# Create nltk_data directory in the user's home directory if it doesn't exist
nltk_data_path = os.path.expanduser("~/nltk_data")
os.makedirs(nltk_data_path, exist_ok=True)
# Add this path to NLTK's data path
nltk.data.path.append(nltk_data_path)
# Download required resources
resources = ['punkt', 'wordnet', 'stopwords', 'punkt_tab']
for resource in resources:
try:
# Different resources can be in different directories in NLTK
locations = [
f'tokenizers/{resource}',
f'corpora/{resource}',
f'taggers/{resource}',
f'{resource}'
]
found = False
for location in locations:
try:
nltk.data.find(location)
print(f"Resource {resource} already downloaded")
found = True
break
except LookupError:
continue
if not found:
print(f"Downloading {resource}...")
nltk.download(resource, quiet=True)
except Exception as e:
print(f"Error with resource {resource}: {e}")
print("NLTK resources check completed")
except Exception as e:
print(f"Error downloading NLTK resources: {e}")
def create_app():
"""
Create a streamlined Gradio app for dataset input and Bag of Words analysis.
Returns:
gr.Blocks: The Gradio application
"""
with gr.Blocks(title="LLM Response Comparator") as app:
# Application state to share data between tabs
dataset_state = gr.State({})
analysis_results_state = gr.State({})
# Dataset Input Tab
with gr.Tab("Dataset Input"):
dataset_inputs, example_dropdown, load_example_btn, create_btn, prompt, response1, model1, response2, model2 = create_dataset_input()
# Add status indicator to show when dataset is created
dataset_status = gr.Markdown("*No dataset loaded*")
# Load example dataset
load_example_btn.click(
fn=load_example_dataset,
inputs=[example_dropdown],
outputs=[prompt, response1, model1, response2, model2] # Update all field values
)
# Save dataset to state and update status
def create_dataset(p, r1, m1, r2, m2):
if not p or not r1 or not r2:
return {}, "❌ **Error:** Please fill in at least the prompt and both responses"
dataset = {
"entries": [
{"prompt": p, "response": r1, "model": m1 or "Model 1"},
{"prompt": p, "response": r2, "model": m2 or "Model 2"}
]
}
return dataset, "✅ **Dataset created successfully!** You can now go to the Analysis tab"
create_btn.click(
fn=create_dataset,
inputs=[prompt, response1, model1, response2, model2],
outputs=[dataset_state, dataset_status]
)
# Analysis Tab
with gr.Tab("Analysis"):
# Use create_analysis_screen to get UI components including visualization container
analysis_options, analysis_params, run_analysis_btn, analysis_output, bow_top_slider, ngram_n, ngram_top, topic_count = create_analysis_screen()
# Pre-create visualization components (initially hidden)
visualization_area_visible = gr.Checkbox(value=False, visible=False, label="Visualization Visible")
analysis_title = gr.Markdown("## Analysis Results", visible=False)
prompt_title = gr.Markdown(visible=False)
models_compared = gr.Markdown(visible=False)
# Container for model 1 words
model1_title = gr.Markdown(visible=False)
model1_words = gr.Markdown(visible=False)
# Container for model 2 words
model2_title = gr.Markdown(visible=False)
model2_words = gr.Markdown(visible=False)
# Similarity metrics
similarity_metrics_title = gr.Markdown("### Similarity Metrics", visible=False)
similarity_metrics = gr.Markdown(visible=False)
# Status or error message area
status_message_visible = gr.Checkbox(value=False, visible=False, label="Status Message Visible")
status_message = gr.Markdown(visible=False)
# Define a helper function to extract parameter values and run the analysis
def run_analysis(dataset, selected_analysis, bow_top, ngram_n, ngram_top, topic_count):
try:
if not dataset or "entries" not in dataset or not dataset["entries"]:
return (
{}, # analysis_results_state
False, # analysis_output visibility
False, # visualization_area_visible
gr.update(visible=False), # analysis_title
gr.update(visible=False), # prompt_title
gr.update(visible=False), # models_compared
gr.update(visible=False), # model1_title
gr.update(visible=False), # model1_words
gr.update(visible=False), # model2_title
gr.update(visible=False), # model2_words
gr.update(visible=False), # similarity_metrics_title
gr.update(visible=False), # similarity_metrics
True, # status_message_visible
gr.update(visible=True, value="❌ **Error:** No dataset loaded. Please create or load a dataset first.") # status_message
)
parameters = {
"bow_top": bow_top,
"ngram_n": ngram_n,
"ngram_top": ngram_top,
"topic_count": topic_count
}
print(f"Running analysis with selected type: {selected_analysis}")
print("Parameters:", parameters)
# Process the analysis request - passing selected_analysis as a string
analysis_results, _ = process_analysis_request(dataset, selected_analysis, parameters)
# If there's an error or no results
if not analysis_results or "analyses" not in analysis_results or not analysis_results["analyses"]:
return (
analysis_results,
False,
False,
gr.update(visible=False),
gr.update(visible=False),
gr.update(visible=False),
gr.update(visible=False),
gr.update(visible=False),
gr.update(visible=False),
gr.update(visible=False),
gr.update(visible=False),
gr.update(visible=False),
True,
gr.update(visible=True, value="❌ **No results found.** Try a different analysis option.")
)
# Extract information to display in components
prompt = list(analysis_results["analyses"].keys())[0]
analyses = analysis_results["analyses"][prompt]
# Initialize visualization components visibilities and contents
visualization_area_visible = False
prompt_title_visible = False
prompt_title_value = ""
models_compared_visible = False
models_compared_value = ""
model1_title_visible = False
model1_title_value = ""
model1_words_visible = False
model1_words_value = ""
model2_title_visible = False
model2_title_value = ""
model2_words_visible = False
model2_words_value = ""
similarity_title_visible = False
similarity_metrics_visible = False
similarity_metrics_value = ""
# Check for messages from placeholder analyses
if "message" in analyses:
return (
analysis_results,
False,
False,
gr.update(visible=False),
gr.update(visible=False),
gr.update(visible=False),
gr.update(visible=False),
gr.update(visible=False),
gr.update(visible=False),
gr.update(visible=False),
gr.update(visible=False),
gr.update(visible=False),
True,
gr.update(visible=True, value=f"ℹ️ **{analyses['message']}**")
)
# Process based on the selected analysis type
if selected_analysis == "Bag of Words" and "bag_of_words" in analyses:
visualization_area_visible = True
bow_results = analyses["bag_of_words"]
models = bow_results.get("models", [])
if len(models) >= 2:
prompt_title_visible = True
prompt_title_value = f"## Analysis of Prompt: \"{prompt[:100]}...\""
models_compared_visible = True
models_compared_value = f"### Comparing responses from {models[0]} and {models[1]}"
# Extract and format information for display
model1_name = models[0]
model2_name = models[1]
# Format important words for each model
important_words = bow_results.get("important_words", {})
if model1_name in important_words:
model1_title_visible = True
model1_title_value = f"#### Top Words Used by {model1_name}"
word_list = [f"**{item['word']}** ({item['count']})" for item in important_words[model1_name][:10]]
model1_words_visible = True
model1_words_value = ", ".join(word_list)
if model2_name in important_words:
model2_title_visible = True
model2_title_value = f"#### Top Words Used by {model2_name}"
word_list = [f"**{item['word']}** ({item['count']})" for item in important_words[model2_name][:10]]
model2_words_visible = True
model2_words_value = ", ".join(word_list)
# Format similarity metrics
comparisons = bow_results.get("comparisons", {})
comparison_key = f"{model1_name} vs {model2_name}"
if comparison_key in comparisons:
metrics = comparisons[comparison_key]
cosine = metrics.get("cosine_similarity", 0)
jaccard = metrics.get("jaccard_similarity", 0)
semantic = metrics.get("semantic_similarity", 0)
common_words = metrics.get("common_word_count", 0)
similarity_title_visible = True
similarity_metrics_visible = True
similarity_metrics_value = f"""
- **Cosine Similarity**: {cosine:.2f} (higher means more similar word frequency patterns)
- **Jaccard Similarity**: {jaccard:.2f} (higher means more word overlap)
- **Semantic Similarity**: {semantic:.2f} (higher means more similar meaning)
- **Common Words**: {common_words} words appear in both responses
"""
# Check for N-gram analysis
elif selected_analysis == "N-gram Analysis" and "ngram_analysis" in analyses:
visualization_area_visible = True
ngram_results = analyses["ngram_analysis"]
models = ngram_results.get("models", [])
ngram_size = ngram_results.get("ngram_size", 2)
size_name = "Unigrams" if ngram_size == 1 else f"{ngram_size}-grams"
if len(models) >= 2:
prompt_title_visible = True
prompt_title_value = f"## Analysis of Prompt: \"{prompt[:100]}...\""
models_compared_visible = True
models_compared_value = f"### {size_name} Analysis: Comparing responses from {models[0]} and {models[1]}"
# Extract and format information for display
model1_name = models[0]
model2_name = models[1]
# Format important n-grams for each model
important_ngrams = ngram_results.get("important_ngrams", {})
if model1_name in important_ngrams:
model1_title_visible = True
model1_title_value = f"#### Top {size_name} Used by {model1_name}"
ngram_list = [f"**{item['ngram']}** ({item['count']})" for item in important_ngrams[model1_name][:10]]
model1_words_visible = True
model1_words_value = ", ".join(ngram_list)
if model2_name in important_ngrams:
model2_title_visible = True
model2_title_value = f"#### Top {size_name} Used by {model2_name}"
ngram_list = [f"**{item['ngram']}** ({item['count']})" for item in important_ngrams[model2_name][:10]]
model2_words_visible = True
model2_words_value = ", ".join(ngram_list)
# Format similarity metrics if available
if "comparisons" in ngram_results:
comparison_key = f"{model1_name} vs {model2_name}"
if comparison_key in ngram_results["comparisons"]:
metrics = ngram_results["comparisons"][comparison_key]
common_count = metrics.get("common_ngram_count", 0)
similarity_title_visible = True
similarity_metrics_visible = True
similarity_metrics_value = f"""
- **Common {size_name}**: {common_count} {size_name.lower()} appear in both responses
"""
# Check for Topic Modeling analysis
elif selected_analysis == "Topic Modeling" and "topic_modeling" in analyses:
visualization_area_visible = True
topic_results = analyses["topic_modeling"]
models = topic_results.get("models", [])
method = topic_results.get("method", "lda").upper()
n_topics = topic_results.get("n_topics", 3)
if len(models) >= 2:
prompt_title_visible = True
prompt_title_value = f"## Analysis of Prompt: \"{prompt[:100]}...\""
models_compared_visible = True
models_compared_value = f"### Topic Modeling Analysis ({method}, {n_topics} topics)"
# Extract and format topic information
topics = topic_results.get("topics", [])
if topics:
# Format topic info for display
topic_info = []
for topic in topics[:3]: # Show first 3 topics
topic_id = topic.get("id", 0)
words = topic.get("words", [])[:5] # Top 5 words per topic
if words:
topic_info.append(f"**Topic {topic_id+1}**: {', '.join(words)}")
if topic_info:
model1_title_visible = True
model1_title_value = "#### Discovered Topics"
model1_words_visible = True
model1_words_value = "\n".join(topic_info)
# Get topic distributions for models
model_topics = topic_results.get("model_topics", {})
if model_topics:
model1_name = models[0]
model2_name = models[1]
# Format topic distribution info
if model1_name in model_topics and model2_name in model_topics:
model2_title_visible = True
model2_title_value = "#### Topic Distribution"
model2_words_visible = True
# Simple distribution display
dist1 = model_topics[model1_name]
dist2 = model_topics[model2_name]
model2_words_value = f"""
**{model1_name}**: {', '.join([f"Topic {i+1}: {v:.2f}" for i, v in enumerate(dist1[:3])])}
**{model2_name}**: {', '.join([f"Topic {i+1}: {v:.2f}" for i, v in enumerate(dist2[:3])])}
"""
# Add similarity metrics if available
comparisons = topic_results.get("comparisons", {})
if comparisons:
comparison_key = f"{model1_name} vs {model2_name}"
if comparison_key in comparisons:
metrics = comparisons[comparison_key]
js_div = metrics.get("js_divergence", 0)
similarity_title_visible = True
similarity_metrics_visible = True
similarity_metrics_value = f"""
- **Topic Distribution Divergence**: {js_div:.4f} (lower means more similar topic distributions)
"""
# Check for Classifier analysis
elif selected_analysis == "Classifier" and "classifier" in analyses:
visualization_area_visible = True
classifier_results = analyses["classifier"]
models = classifier_results.get("models", [])
if len(models) >= 2:
prompt_title_visible = True
prompt_title_value = f"## Analysis of Prompt: \"{prompt[:100]}...\""
models_compared_visible = True
models_compared_value = f"### Classifier Analysis for {models[0]} and {models[1]}"
# Extract and format classifier information
model1_name = models[0]
model2_name = models[1]
# Display classifications for each model
classifications = classifier_results.get("classifications", {})
if classifications:
model1_title_visible = True
model1_title_value = f"#### Classification Results"
model1_words_visible = True
model1_results = classifications.get(model1_name, {})
model2_results = classifications.get(model2_name, {})
model1_words_value = f"""
**{model1_name}**:
- Formality: {model1_results.get('formality', 'N/A')}
- Sentiment: {model1_results.get('sentiment', 'N/A')}
- Complexity: {model1_results.get('complexity', 'N/A')}
**{model2_name}**:
- Formality: {model2_results.get('formality', 'N/A')}
- Sentiment: {model2_results.get('sentiment', 'N/A')}
- Complexity: {model2_results.get('complexity', 'N/A')}
"""
# Show comparison
model2_title_visible = True
model2_title_value = f"#### Classification Comparison"
model2_words_visible = True
differences = classifier_results.get("differences", {})
model2_words_value = "\n".join([
f"- **{category}**: {diff}"
for category, diff in differences.items()
])
# If we don't have visualization data from any analysis
if not visualization_area_visible:
return (
analysis_results,
False,
False,
gr.update(visible=False),
gr.update(visible=False),
gr.update(visible=False),
gr.update(visible=False),
gr.update(visible=False),
gr.update(visible=False),
gr.update(visible=False),
gr.update(visible=False),
gr.update(visible=False),
True,
gr.update(visible=True, value="❌ **No visualization data found.** Make sure to select a valid analysis option.")
)
# Return all updated component values
return (
analysis_results, # analysis_results_state
False, # analysis_output visibility
True, # visualization_area_visible
gr.update(visible=True), # analysis_title
gr.update(visible=prompt_title_visible, value=prompt_title_value), # prompt_title
gr.update(visible=models_compared_visible, value=models_compared_value), # models_compared
gr.update(visible=model1_title_visible, value=model1_title_value), # model1_title
gr.update(visible=model1_words_visible, value=model1_words_value), # model1_words
gr.update(visible=model2_title_visible, value=model2_title_value), # model2_title
gr.update(visible=model2_words_visible, value=model2_words_value), # model2_words
gr.update(visible=similarity_title_visible), # similarity_metrics_title
gr.update(visible=similarity_metrics_visible, value=similarity_metrics_value), # similarity_metrics
False, # status_message_visible
gr.update(visible=False) # status_message
)
except Exception as e:
import traceback
error_msg = f"Error in analysis: {str(e)}\n{traceback.format_exc()}"
print(error_msg)
return (
{"error": error_msg}, # analysis_results_state
True, # analysis_output visibility (show raw JSON for debugging)
False, # visualization_area_visible
gr.update(visible=False),
gr.update(visible=False),
gr.update(visible=False),
gr.update(visible=False),
gr.update(visible=False),
gr.update(visible=False),
gr.update(visible=False),
gr.update(visible=False),
gr.update(visible=False),
True, # status_message_visible
gr.update(visible=True, value=f"❌ **Error during analysis:**\n\n```\n{str(e)}\n```") # status_message
)
# Add a new LLM Analysis tab
with gr.Tab("LLM Analysis"):
gr.Markdown("## LLM-Based Response Analysis")
with gr.Row():
with gr.Column():
llm_analysis_type = gr.Radio(
choices=["Response Quality", "Response Comparison", "Factual Accuracy"],
label="Analysis Type",
value="Response Comparison"
)
llm_model = gr.Dropdown(
choices=["OpenAI GPT-4", "Anthropic Claude", "Local LLM"],
label="Analysis Model",
value="OpenAI GPT-4"
)
run_llm_analysis_btn = gr.Button("Run LLM Analysis", variant="primary")
with gr.Column():
llm_analysis_prompt = gr.Textbox(
label="Custom Analysis Instructions (Optional)",
placeholder="Enter any specific instructions for the analysis...",
lines=3
)
llm_analysis_status = gr.Markdown("*No analysis has been run*")
llm_analysis_result = gr.Markdown(visible=False)
# Placeholder function for LLM analysis
def run_llm_analysis(dataset, analysis_type, model, custom_prompt):
if not dataset or "entries" not in dataset or not dataset["entries"]:
return (
gr.update(visible=True, value="❌ **Error:** No dataset loaded. Please create or load a dataset first."),
gr.update(visible=False)
)
# Placeholder for actual implementation
return (
gr.update(visible=True, value="⏳ **Implementation in progress**\n\nLLM-based analysis will be available in a future update."),
gr.update(visible=False)
)
# Connect the run button to the analysis function
run_llm_analysis_btn.click(
fn=run_llm_analysis,
inputs=[dataset_state, llm_analysis_type, llm_model, llm_analysis_prompt],
outputs=[llm_analysis_status, llm_analysis_result]
)
# Run analysis with proper parameters
run_analysis_btn.click(
fn=run_analysis,
inputs=[dataset_state, analysis_options, bow_top_slider, ngram_n, ngram_top, topic_count],
outputs=[
analysis_results_state,
analysis_output,
visualization_area_visible,
analysis_title,
prompt_title,
models_compared,
model1_title,
model1_words,
model2_title,
model2_words,
similarity_metrics_title,
similarity_metrics,
status_message_visible,
status_message
]
)
return app
if __name__ == "__main__":
# Download required NLTK resources before launching the app
download_nltk_resources()
app = create_app()
app.launch() |