Generate training data and FAQ pairs for customer support chatbot systems.
## Chatbot FAQ Trainer Prompt Generate training data for a customer support chatbot: **Existing FAQs:** [EXISTING_FAQS] **Conversation Logs:** [CONVERSATION_LOGS] **Product:** [PRODUCT_NAME] **Training Data Generation:** ### 1. Intent Classification Pairs For each FAQ, generate 10 variant phrasings: **Intent:** [INTENT_NAME] **Canonical Question:** [FAQ_QUESTION] **Variants:** 1. [Casual phrasing] 2. [Formal phrasing] 3. [Misspelled version] 4. [Abbreviated version] 5. [Question with context] 6. [Angry phrasing] 7. [Confused phrasing] 8. [Mobile/short phrasing] 9. [Non-native English phrasing] 10. [Voice-to-text phrasing with errors] **Answer:** [FAQ_ANSWER] **Follow-up Questions:** [LIKELY_FOLLOW_UPS] ### 2. Conversation Flow Training Generate 5 multi-turn conversations per intent: ``` User: [Opening message] Bot: [Response] User: [Follow-up] Bot: [Response] User: [Clarification or new topic] Bot: [Appropriate transition] ``` ### 3. Edge Cases - Ambiguous queries that could match multiple intents - Out-of-scope questions the bot should redirect - Emotional/angry messages requiring empathy - Multi-intent messages (asking about 2+ things) ### 4. Fallback Training - Phrases that should trigger human handoff - Responses for "I don't understand" scenarios - Graceful degradation paths **Quality Metrics:** - Intent recognition accuracy target: [X]% - False positive rate target: < [X]% - Fallback trigger rate target: < [X]% - Customer satisfaction with bot: [X]/5
Free to copy and use. Compatible with Claude 4 Opus, GPT-5, Llama 4.
Provide existing FAQ content in [EXISTING_FAQS] and conversation logs.
Initial release
Sign in and download this prompt to leave a review.