halukakin / ai-check-conv-llama3-8b

**Task:** For conversations in Turkish judge conversation safety. If there are any underage participants. **Criteria:** * Speakers under 18, if any of the users claim to be under 18 years of age * Suspicious sentences, if any of the users ask for money, video calls, or any other inappropriate requests * If the conversation is harmless, return false * If the conversation contains speakers under 18, return true and include the suspicious sentence * If the conversation contains inappropriate requests, return true and include the suspicious sentence * If the conversation contains both underage speakers and inappropriate requests, return true and include the suspicious sentence **Output:** * True: Conversation contains underage speakers or someone asks for money, or any other inappropriate requests. * Include "suspicious_sentence" if applicable. * False: Conversation is harmless. Response should be in JSON format. **Negative Messages (Examples):** * Conversations involving participants under 18 * Conversations involving inappropriate requests * Conversations involving both participants under 18 and inappropriate requests * Conversations involve money requests * Conversations involve loan requests * Conversations involve video call requests * Conversations involve inappropriate requests * Conversations involve suspicious sentences * Conversations involve webcam show recommendations * Conversations involve credit card or IBAN number requests **Example Response (True):** ```json { "is_inappropriate": true, "suspicious_sentence": "<Suspicious Sentence>" } **Example Response (False):** ```json { "is_inappropriate": false, "suspicious_sentence": "<NONE>" }