Eureka-Leo's picture
Add files using upload-large-folder tool
ba1d61a verified
import json
import re
import ast
from sklearn.metrics import accuracy_score, f1_score, classification_report, confusion_matrix
from collections import Counter, defaultdict
from datetime import datetime
import numpy as np
def extract_emotion_intent_from_output(model_output):
if not model_output:
return None, None, False, "empty_output"
valid_emotions = ['happy', 'surprise', 'sad', 'disgust', 'anger', 'fear', 'neutral']
valid_intents = ['questioning', 'agreeing', 'acknowledging', 'encouraging',
'consoling', 'suggesting', 'wishing', 'neutral']
try:
if "{'emotion':" in model_output or '{"emotion":' in model_output:
cleaned_output = model_output.strip()
cleaned_output = re.sub(r"'intent'(\s*,\s*)'", r"'intent':\1'", cleaned_output)
cleaned_output = re.sub(r'"intent"(\s*,\s*)"', r'"intent":\1"', cleaned_output)
if not cleaned_output.startswith('{'):
json_match = re.search(r'\{[^}]*\}', cleaned_output)
if json_match:
cleaned_output = json_match.group()
try:
parsed = ast.literal_eval(cleaned_output)
except:
parsed = json.loads(cleaned_output)
if 'emotion' in parsed and 'intent' in parsed:
emotion = parsed['emotion'].lower().strip()
intent = parsed['intent'].lower().strip()
emotion_valid = emotion in valid_emotions
intent_valid = intent in valid_intents
if emotion_valid and intent_valid:
return emotion, intent, True, None
elif not emotion_valid and not intent_valid:
return emotion, intent, False, "invalid_emotion_and_intent"
elif not emotion_valid:
return emotion, intent, False, "invalid_emotion"
else:
return emotion, intent, False, "invalid_intent"
emotion_found = None
intent_found = None
for emotion in valid_emotions:
if emotion in model_output.lower():
emotion_found = emotion
break
for intent in valid_intents:
if intent in model_output.lower():
intent_found = intent
break
if emotion_found and intent_found:
return emotion_found, intent_found, False, "labels_found_but_not_properly_formatted"
elif emotion_found:
return emotion_found, None, False, "only_emotion_found"
elif intent_found:
return None, intent_found, False, "only_intent_found"
return None, None, False, "no_labels_pattern"
except Exception as e:
return None, None, False, f"parsing_error_{str(e)}"
def parse_ground_truth(ground_truth):
if isinstance(ground_truth, dict):
return ground_truth.get('emotion', '').lower(), ground_truth.get('intent', '').lower()
elif isinstance(ground_truth, str):
cleaned_gt = ground_truth.strip()
cleaned_gt = re.sub(r"'intent'(\s*,\s*)'", r"'intent':\1'", cleaned_gt)
try:
parsed = ast.literal_eval(cleaned_gt)
return parsed.get('emotion', '').lower(), parsed.get('intent', '').lower()
except:
return '', ''
else:
return '', ''
def evaluate_emotion_based_intent_analysis(result_file_path):
with open(result_file_path, 'r', encoding='utf-8') as f:
results = json.load(f)
emotion_predictions = []
emotion_ground_truths = []
intent_predictions = []
intent_ground_truths = []
detailed_results = []
extraction_errors = defaultdict(list)
prediction_errors = defaultdict(list)
emotion_labels = ['happy', 'surprise', 'sad', 'disgust', 'anger', 'fear', 'neutral']
intent_labels = ['questioning', 'agreeing', 'acknowledging', 'encouraging',
'consoling', 'suggesting', 'wishing', 'neutral']
for item in results:
item_id = item['id']
model_output = item['model_output']
gt_emotion, gt_intent = parse_ground_truth(item['ground_truth'])
pred_emotion, pred_intent, is_valid, error_type = extract_emotion_intent_from_output(model_output)
detailed_item = {
'id': item_id,
'model_output': model_output,
'extracted_prediction': {
'emotion': pred_emotion,
'intent': pred_intent
},
'ground_truth': {
'emotion': gt_emotion,
'intent': gt_intent
},
'emotion_correct': pred_emotion == gt_emotion if pred_emotion else False,
'intent_correct': pred_intent == gt_intent if pred_intent else False,
'both_correct': (pred_emotion == gt_emotion and pred_intent == gt_intent) if (pred_emotion and pred_intent) else False,
'valid': is_valid
}
detailed_results.append(detailed_item)
if not is_valid:
extraction_errors[error_type].append(item_id)
else:
if pred_emotion != gt_emotion:
error_pattern = f"emotion_{gt_emotion}_to_{pred_emotion}"
prediction_errors[error_pattern].append(item_id)
if pred_intent != gt_intent:
error_pattern = f"intent_{gt_intent}_to_{pred_intent}"
prediction_errors[error_pattern].append(item_id)
if is_valid:
emotion_predictions.append(pred_emotion)
emotion_ground_truths.append(gt_emotion)
intent_predictions.append(pred_intent)
intent_ground_truths.append(gt_intent)
if len(emotion_predictions) == 0:
return {
'error': 'No valid predictions found',
'total_samples': len(results),
'extraction_errors': dict(extraction_errors)
}
emotion_accuracy = accuracy_score(emotion_ground_truths, emotion_predictions)
emotion_weighted_f1 = f1_score(emotion_ground_truths, emotion_predictions, average='weighted')
emotion_macro_f1 = f1_score(emotion_ground_truths, emotion_predictions, average='macro')
intent_accuracy = accuracy_score(intent_ground_truths, intent_predictions)
intent_weighted_f1 = f1_score(intent_ground_truths, intent_predictions, average='weighted')
intent_macro_f1 = f1_score(intent_ground_truths, intent_predictions, average='macro')
both_correct = sum(1 for item in detailed_results if item['both_correct'])
joint_accuracy = both_correct / len([item for item in detailed_results if item['valid']])
emotion_cm = confusion_matrix(emotion_ground_truths, emotion_predictions, labels=emotion_labels)
intent_cm = confusion_matrix(intent_ground_truths, intent_predictions, labels=intent_labels)
emotion_class_report = classification_report(emotion_ground_truths, emotion_predictions,
target_names=emotion_labels,
output_dict=True, zero_division=0)
intent_class_report = classification_report(intent_ground_truths, intent_predictions,
target_names=intent_labels,
output_dict=True, zero_division=0)
evaluation_result = {
'task_info': {
'task_name': 'emotion.based.intent.analysis',
'dataset': 'MC-EIU',
'evaluation_time': datetime.now().isoformat(),
'total_samples': len(results),
'valid_predictions': len(emotion_predictions),
'extraction_success_rate': round(len(emotion_predictions) / len(results), 4)
},
'metrics': {
'emotion_metrics': {
'ACC': round(emotion_accuracy, 4),
'WAF': round(emotion_weighted_f1, 4),
'Macro_F1': round(emotion_macro_f1, 4)
},
'intent_metrics': {
'ACC': round(intent_accuracy, 4),
'WAF': round(intent_weighted_f1, 4),
'Macro_F1': round(intent_macro_f1, 4)
},
'joint_metrics': {
'Joint_ACC': round(joint_accuracy, 4)
}
},
'per_class_metrics': {
'emotion': {
label: {
'precision': round(emotion_class_report[label]['precision'], 4),
'recall': round(emotion_class_report[label]['recall'], 4),
'f1_score': round(emotion_class_report[label]['f1-score'], 4),
'support': int(emotion_class_report[label]['support'])
} for label in emotion_labels if label in emotion_class_report
},
'intent': {
label: {
'precision': round(intent_class_report[label]['precision'], 4),
'recall': round(intent_class_report[label]['recall'], 4),
'f1_score': round(intent_class_report[label]['f1-score'], 4),
'support': int(intent_class_report[label]['support'])
} for label in intent_labels if label in intent_class_report
}
},
'confusion_matrices': {
'emotion': {
'labels': emotion_labels,
'matrix': emotion_cm.tolist()
},
'intent': {
'labels': intent_labels,
'matrix': intent_cm.tolist()
}
},
'error_analysis': {
'extraction_errors': {
error_type: {
'count': len(sample_ids),
'sample_ids': sample_ids
} for error_type, sample_ids in extraction_errors.items()
},
'prediction_errors': {
error_pattern: {
'count': len(sample_ids),
'sample_ids': sample_ids
} for error_pattern, sample_ids in prediction_errors.items()
}
},
'distribution': {
'emotion': {
'ground_truth': dict(Counter(emotion_ground_truths)),
'predictions': dict(Counter(emotion_predictions))
},
'intent': {
'ground_truth': dict(Counter(intent_ground_truths)),
'predictions': dict(Counter(intent_predictions))
}
}
}
base_name = result_file_path.replace('.json', '')
eval_output_file = f"{base_name}_evaluation.json"
with open(eval_output_file, 'w', encoding='utf-8') as f:
json.dump(evaluation_result, f, ensure_ascii=False, indent=2)
detailed_output_file = f"{base_name}_detailed_results.json"
with open(detailed_output_file, 'w', encoding='utf-8') as f:
json.dump(detailed_results, f, ensure_ascii=False, indent=2)
problem_samples = [item for item in detailed_results if not item['both_correct']]
if problem_samples:
problem_report_file = f"{base_name}_problem_samples.json"
with open(problem_report_file, 'w', encoding='utf-8') as f:
json.dump(problem_samples, f, ensure_ascii=False, indent=2)
print(f"Evaluation complete: {len(results)} samples")
print(f"Emotion metrics: ACC={evaluation_result['metrics']['emotion_metrics']['ACC']}, WAF={evaluation_result['metrics']['emotion_metrics']['WAF']}")
print(f"Intent metrics: ACC={evaluation_result['metrics']['intent_metrics']['ACC']}, WAF={evaluation_result['metrics']['intent_metrics']['WAF']}")
print(f"Joint accuracy: {evaluation_result['metrics']['joint_metrics']['Joint_ACC']}")
print(f"Extraction success rate: {evaluation_result['task_info']['extraction_success_rate']}")
print(f"Results saved to: {eval_output_file}")
if problem_samples:
print(f"Problematic samples: {len(problem_samples)}; see {problem_report_file} for details")
return evaluation_result
if __name__ == "__main__":
result_file = "model_result.json"
try:
evaluation_result = evaluate_emotion_based_intent_analysis(result_file)
except FileNotFoundError:
print(f"Error: file not found {result_file}")
except json.JSONDecodeError:
print(f"Error: invalid format for {result_file}")
except Exception as e:
print(f"Evaluation failed: {str(e)}")