Relevance Tuning

def hybrid_search(query, keyword_results, semantic_results, user_context):
    # Get base scores
    keyword_scores = [score for doc, score in keyword_results]
    semantic_scores = [score for doc, score in semantic_results]

    # Normalize scores
    keyword_scores = normalize(keyword_scores)
    semantic_scores = normalize(semantic_scores)

    # Predict optimal weights based on query and user context
    keyword_weight, semantic_weight = predict_weights(query, user_context)

    # Combine scores
    final_scores = [
        keyword_weight * k_score + semantic_weight * s_score
        for k_score, s_score in zip(keyword_scores, semantic_scores)
    ]

    # Sort and return results
    combined_results = list(zip(keyword_results, semantic_results, final_scores))
    return sorted(combined_results, key=lambda x: x[2], reverse=True)

def predict_weights(query, user_context):
    # Use a machine learning model to predict optimal weights
    # based on query characteristics and user context
    # This is a placeholder for the actual implementation
    return 0.6, 0.4  # Example weights