Questions Refinement

You have generated a list of questions and their contexts, based on a document that was provided to you. Some questions are simple, direct questions, and the others are complex questions. Each question is associated with relevant context from the document.

1.Analyze: Carefully compare each question and its context 2.Identify Overlap: Determine if any questions ask for essentially the same information or have a very similar intent, even if phrased differently. Consider these factors:

Exact Question Match: Are the questions worded identically? Contextual Overlap: Do the contexts of the questions significantly overlap?

Semantic Similarity: Do the questions have very similar meanings despite different wordings?

Prioritize Simple Questions: If a complex question and a simple question are identified as overlapping, always retain the simple question in favor of the complex question.

4. Output: Compile the final list of questions and the associated contexts in the language of the questions, as a JSON dictionary as an array of objects within a "questions" key, each containing "question" and "contexts" properties. Do not include any commentary or extraneous text

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

Example 1: Overlapping Removal * Simple Question: What is the capital of France? * Overlapping Complex Question: Investigate the historical significance of Paris as the capital of France. * This complex question would be removed as it only seeks the same information as the simple one. Do not include commentary or extraneous text

Questions: {questions}