Multitask Learning for Crash Analysis: A Fine-Tuned LLM Framework Using Twitter Data

Keyphrases

• Crash Analysis 100%
• Information Retrieval 100%
• Multi-task Learning 100%
• Large Language Models 100%
• Road Traffic Crashes 100%
• Classification Task 66%
• Multi-task Learning Framework 66%
• GPT2 Model 66%
• Mini-model 66%
• IScore 66%
• Retrieval Task 66%
• XGBoost 66%
• ROUGE 66%
• Traffic Safety 33%
• Natural Language 33%
• Language Modeling 33%
• Traffic Safety Analysis 33%
• GPT-2 33%
• Multi-task 33%
• Road Safety 33%
• Task Classifier 33%
• Zero-shot 33%
• Mini 33%
• GPT-4 33%
• GPT-3.5 33%
• Data Delay 33%
• Data-driven Decision Making 33%
• Incomplete Data 33%
• Real-time Traffic 33%
• Traffic-related 33%

Computer Science

• Multitask Learning 100%
• Large Language Model 100%
• Information Retrieval 75%
• Extreme Gradient Boosting 50%
• Learning Framework 50%
• Classification Task 50%
• Generative Pre-Trained Transformer 4 25%
• Average Accuracy 25%
• Language Modeling 25%
• Baseline Model 25%
• Learning Approach 25%

Neuroscience

• Alanine Aminotransferase 100%
• Decision-Making 14%