25,715 non-empty text examples categorized into 68 unique intents and 18 distinct scenarios. The collection provides raw utterance data for benchmarking Natural Language Understanding systems across diverse conversational domains.
Use Cases
- Train intent recognition models using the 68 unique intent labels and corresponding text utterances
- Perform domain-specific performance analysis by grouping results according to the 18 scenario categories
- Develop hierarchical classification systems that first predict the scenario and then the specific intent
Strengths
- 25,715 non-empty text examples for NLU benchmarking
- 68 unique intent labels for fine-grained classification tasks
- 18 scenario categories providing high-level context for the intent data