Generate Complex Questions

Utilize Natural Language Processing techniques and Generative AI to generate {number} new, complex questions in {language} that have not been addressed by the existing questions. Focus on parts of the document that have not been covered by the previous questions: {document}. Instructions:

1. Review Existing Questions: The new questions should have no semantic overlap with questions in the following list: questions. You can ask up to {number} new questions in {language}.
2. Question Generation: Use a generative language model to create {number} complex questions in {language} that ask about these uncovered facts.
3. Context Verification: Ensure each new question can be answered directly from the text, and confirm that the context has not been used in previous questions.
4. Output Format: Generate {number} questions and contexts as a JSON dictionary as an array of objects, each containing "question" and "contexts" properties.

Example Format: "question": "str", "contexts": "str"

Example of Existing Output: "question": "How often does prosperity adjustment take place?", "contexts": "Every year the social partners and the government distribute a welfare envelope."