[
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "1",
    "paper_title": "Mapping Essential Urban Land Use Categories in Beijing with a Fast Area of Interest (AOI)-Based Method",
    "question_id": "Q1",
    "question": "What is the number of land-cover / land-use classes classified in this study?",
    "choices": {
      "A": "3",
      "B": "6",
      "C": "9",
      "D": "10",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "D",
    "metadata": {
      "Task-oriented Category": "Study subject & experimental setup",
      "question_key_term": "land cover",
      "term_explanation": "The total number of land cover or land use categories defined for classification. This reflects the complexity of the classification task and determines how fine-grained the land surface representation is. Typical classes include urban, vegetation, water, bare land, etc."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "1",
    "paper_title": "Mapping Essential Urban Land Use Categories in Beijing with a Fast Area of Interest (AOI)-Based Method",
    "question_id": "Q2",
    "question": "What is the spatial extent of the study area?",
    "choices": {
      "A": "16,411 km²",
      "B": "26,035 km²",
      "C": "200,000 km²",
      "D": "1,419,530 km²",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "A",
    "metadata": {
      "Task-oriented Category": "Study subject & experimental setup",
      "question_key_term": "geographic area",
      "term_explanation": "The geographic area covered by the study, typically measured in square kilometers. This affects the generalizability and computational complexity of the study and can range from local sites to national or global scales."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "1",
    "paper_title": "Mapping Essential Urban Land Use Categories in Beijing with a Fast Area of Interest (AOI)-Based Method",
    "question_id": "Q3",
    "question": "What is the geographic type of the study area?",
    "choices": {
      "A": "Urban",
      "B": "Rural",
      "C": "Mixed",
      "D": "Natural (e.g., forest, wetland, desert)",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "A",
    "metadata": {
      "Task-oriented Category": "Study subject & experimental setup",
      "question_key_term": "geographic type",
      "term_explanation": "The  geographic type or functional nature of the study region. Common categories include urban, rural, natural ecosystems (e.g., forest, wetland), or mixed types. The geographic context influences the spectral heterogeneity and classification difficulty."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "1",
    "paper_title": "Mapping Essential Urban Land Use Categories in Beijing with a Fast Area of Interest (AOI)-Based Method",
    "question_id": "Q4",
    "question": "What is the temporal scope of the data used?",
    "choices": {
      "A": "Single-scene imagery",
      "B": "Short-term imagery ( ≤3 months)",
      "C": "Mid-term imagery ( >3 and ≤12 months)",
      "D": "Long-term imagery ( >1 year)",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "C",
    "metadata": {
      "Task-oriented Category": "Data characteristics & collection",
      "question_key_term": "time span",
      "term_explanation": "The total time span covered by the remote sensing imagery used in the study. This could be a single image, a short-term period (weeks to months), seasonal cycles, or long-term historical records (multi-year time series)."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "1",
    "paper_title": "Mapping Essential Urban Land Use Categories in Beijing with a Fast Area of Interest (AOI)-Based Method",
    "question_id": "Q5",
    "question": "What type of remote sensing data is used?",
    "choices": {
      "A": "Optical",
      "B": "SAR",
      "C": "LiDAR",
      "D": "Multisource fusion",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "A",
    "metadata": {
      "Task-oriented Category": "Data characteristics & collection",
      "question_key_term": "data type",
      "term_explanation": "The primary data modality employed in the study. Categories include optical imagery, synthetic aperture radar (SAR), LiDAR, or combinations of multiple sources. The data type determines the spectral, spatial, and structural information available for classification."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "1",
    "paper_title": "Mapping Essential Urban Land Use Categories in Beijing with a Fast Area of Interest (AOI)-Based Method",
    "question_id": "Q6",
    "question": "Which specific satellite data is used?",
    "choices": {
      "A": "Sentinel-2",
      "B": "Sentinel-1",
      "C": "Luojia-1",
      "D": "Sentinel-2 and Luojia-1",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "D",
    "metadata": {
      "Task-oriented Category": "Data characteristics & collection",
      "question_key_term": "satellite",
      "term_explanation": "The name(s) of satellite missions providing the source imagery, such as Sentinel-2, Landsat-8, Sentinel-1, MODIS, or commercial platforms like Planet or WorldView. This helps determine resolution, revisit frequency, and sensor characteristics."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "1",
    "paper_title": "Mapping Essential Urban Land Use Categories in Beijing with a Fast Area of Interest (AOI)-Based Method",
    "question_id": "Q7",
    "question": "What is the spatial resolution of the primary imagery used?",
    "choices": {
      "A": "10 m",
      "B": "16 m",
      "C": "27 m",
      "D": "1000 m",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "A",
    "metadata": {
      "Task-oriented Category": "Data characteristics & collection",
      "question_key_term": "spatial resolution",
      "term_explanation": "The ground sampling distance (in meters) of the input imagery, indicating the size of the smallest detectable object. High-resolution imagery (e.g., <5 m) allows fine-scale mapping, while moderate to coarse resolutions (e.g., 30 m or >100 m) are suited for regional to global scales."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "1",
    "paper_title": "Mapping Essential Urban Land Use Categories in Beijing with a Fast Area of Interest (AOI)-Based Method",
    "question_id": "Q8",
    "question": "Are auxiliary features used beyond raw spectral bands?",
    "choices": {
      "A": "Vegetation indices (e.g., NDVI)",
      "B": "Water features (e.g., NDWI)",
      "C": "Vegetation indices and Water indices",
      "D": "Elevation / DEM",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "C",
    "metadata": {
      "Task-oriented Category": "Data characteristics & collection",
      "question_key_term": "features",
      "term_explanation": "Whether the study utilizes additional derived features such as vegetation indices (NDVI, LAI, FAPAR), land surface temperature, albedo, emissivity, topographic information (e.g., DEM), or other external variables. These features can significantly enhance classification performance."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "1",
    "paper_title": "Mapping Essential Urban Land Use Categories in Beijing with a Fast Area of Interest (AOI)-Based Method",
    "question_id": "Q9",
    "question": "What type of model is implemented in this study?",
    "choices": {
      "A": "SVM",
      "B": "RF",
      "C": "XGBoost",
      "D": "CNN",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "B",
    "metadata": {
      "Task-oriented Category": "Technical approach & details",
      "question_key_term": "ML model",
      "term_explanation": "The classification model used to perform land cover or land use mapping. Options include traditional machine learning algorithms (e.g., random forest, SVM), deep learning architectures (e.g., CNNs, U-Net), or hybrid and transformer-based models. The model choice reflects the learning strategy and computational design."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "1",
    "paper_title": "Mapping Essential Urban Land Use Categories in Beijing with a Fast Area of Interest (AOI)-Based Method",
    "question_id": "Q10",
    "question": "What performance metrics are reported?",
    "choices": {
      "A": "Overall Accuracy (OA)",
      "B": "F1-score",
      "C": "Kappa",
      "D": "OA and Kappa",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "D",
    "metadata": {
      "Task-oriented Category": "Technical approach & details",
      "question_key_term": "performance metrics",
      "term_explanation": "The quantitative indicators used to evaluate model performance, such as overall accuracy (OA), F1-score, Kappa coefficient, and mean intersection-over-union (mIoU). The selection of metrics affects the interpretability and comparability of results."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "1",
    "paper_title": "Mapping Essential Urban Land Use Categories in Beijing with a Fast Area of Interest (AOI)-Based Method",
    "question_id": "Q11",
    "question": "Is any comparative analysis included?",
    "choices": {
      "A": "Compared with traditional classifiers (e.g., RF, SVM)",
      "B": "Compared with deep models (e.g., U-Net variants)",
      "C": "Ablation study",
      "D": "No comparison reported",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "D",
    "metadata": {
      "Task-oriented Category": "Conclusions & results",
      "question_key_term": "comparative analysis",
      "term_explanation": "Whether the study conducts comparisons with baseline models, traditional classifiers, or variants of the proposed method (e.g., ablation study, benchmark comparison). This provides insight into model robustness and improvement over existing methods."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "1",
    "paper_title": "Mapping Essential Urban Land Use Categories in Beijing with a Fast Area of Interest (AOI)-Based Method",
    "question_id": "Q12",
    "question": "What is the reported overall accuracy (OA)?",
    "choices": {
      "A": "69%",
      "B": "74%",
      "C": "77%",
      "D": "98%",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "C",
    "metadata": {
      "Task-oriented Category": "Conclusions & results",
      "question_key_term": "overall accuracy",
      "term_explanation": "The percentage of correctly classified samples over the total number of samples, as reported in the study. OA is a commonly used metric for remote sensing classification and provides a general measure of model performance."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "2",
    "paper_title": "Mapping essential urban land use categories in China (EULUC-China): preliminary results for 2018",
    "question_id": "Q1",
    "question": "What is the number of land-cover / land-use classes classified in this study?",
    "choices": {
      "A": "5",
      "B": "12",
      "C": "21",
      "D": "37",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "B",
    "metadata": {
      "Task-oriented Category": "Study subject & experimental setup",
      "question_key_term": "land cover",
      "term_explanation": "The total number of land cover or land use categories defined for classification. This reflects the complexity of the classification task and determines how fine-grained the land surface representation is. Typical classes include urban, vegetation, water, bare land, etc."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "2",
    "paper_title": "Mapping essential urban land use categories in China (EULUC-China): preliminary results for 2018",
    "question_id": "Q2",
    "question": "What is the spatial extent of the study area?",
    "choices": {
      "A": "7,317 km²",
      "B": "41,576 km²",
      "C": "67,558 km²",
      "D": "166,338 km²",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "D",
    "metadata": {
      "Task-oriented Category": "Study subject & experimental setup",
      "question_key_term": "geographic area",
      "term_explanation": "The geographic area covered by the study, typically measured in square kilometers. This affects the generalizability and computational complexity of the study and can range from local sites to national or global scales."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "2",
    "paper_title": "Mapping essential urban land use categories in China (EULUC-China): preliminary results for 2018",
    "question_id": "Q3",
    "question": "What is the geographic type of the study area?",
    "choices": {
      "A": "Urban",
      "B": "Rural",
      "C": "Mixed",
      "D": "Natural (e.g., forest, wetland, desert)",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "A",
    "metadata": {
      "Task-oriented Category": "Study subject & experimental setup",
      "question_key_term": "geographic type",
      "term_explanation": "The  geographic type or functional nature of the study region. Common categories include urban, rural, natural ecosystems (e.g., forest, wetland), or mixed types. The geographic context influences the spectral heterogeneity and classification difficulty."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "2",
    "paper_title": "Mapping essential urban land use categories in China (EULUC-China): preliminary results for 2018",
    "question_id": "Q4",
    "question": "What is the temporal scope of the data used?",
    "choices": {
      "A": "Single-scene imagery",
      "B": "Short-term imagery ( ≤3 months)",
      "C": "Mid-term imagery ( >3 and ≤12 months)",
      "D": "Long-term imagery ( >1 year)",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "C",
    "metadata": {
      "Task-oriented Category": "Data characteristics & collection",
      "question_key_term": "time span",
      "term_explanation": "The total time span covered by the remote sensing imagery used in the study. This could be a single image, a short-term period (weeks to months), seasonal cycles, or long-term historical records (multi-year time series)."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "2",
    "paper_title": "Mapping essential urban land use categories in China (EULUC-China): preliminary results for 2018",
    "question_id": "Q5",
    "question": "What type of remote sensing data is used?",
    "choices": {
      "A": "Optical",
      "B": "SAR",
      "C": "LiDAR",
      "D": "Multisource fusion",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "A",
    "metadata": {
      "Task-oriented Category": "Data characteristics & collection",
      "question_key_term": "data type",
      "term_explanation": "The primary data modality employed in the study. Categories include optical imagery, synthetic aperture radar (SAR), LiDAR, or combinations of multiple sources. The data type determines the spectral, spatial, and structural information available for classification."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "2",
    "paper_title": "Mapping essential urban land use categories in China (EULUC-China): preliminary results for 2018",
    "question_id": "Q6",
    "question": "Which specific satellite data is used?",
    "choices": {
      "A": "Sentinel-1",
      "B": "Sentinel-2",
      "C": "Luojia-1",
      "D": "Sentinel-2 and Luojia-1",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "D",
    "metadata": {
      "Task-oriented Category": "Data characteristics & collection",
      "question_key_term": "satellite",
      "term_explanation": "The name(s) of satellite missions providing the source imagery, such as Sentinel-2, Landsat-8, Sentinel-1, MODIS, or commercial platforms like Planet or WorldView. This helps determine resolution, revisit frequency, and sensor characteristics."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "2",
    "paper_title": "Mapping essential urban land use categories in China (EULUC-China): preliminary results for 2018",
    "question_id": "Q7",
    "question": "What is the spatial resolution of the primary imagery used?",
    "choices": {
      "A": "2 m",
      "B": "10 m",
      "C": "21 m",
      "D": "27 m",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "B",
    "metadata": {
      "Task-oriented Category": "Data characteristics & collection",
      "question_key_term": "spatial resolution",
      "term_explanation": "The ground sampling distance (in meters) of the input imagery, indicating the size of the smallest detectable object. High-resolution imagery (e.g., <5 m) allows fine-scale mapping, while moderate to coarse resolutions (e.g., 30 m or >100 m) are suited for regional to global scales."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "2",
    "paper_title": "Mapping essential urban land use categories in China (EULUC-China): preliminary results for 2018",
    "question_id": "Q8",
    "question": "Are auxiliary features used beyond raw spectral bands?",
    "choices": {
      "A": "Vegetation indices (e.g., NDVI)",
      "B": "Water features (e.g., NDWI)",
      "C": "Vegetation indices and Water indices",
      "D": "Elevation / DEM",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "D",
    "metadata": {
      "Task-oriented Category": "Data characteristics & collection",
      "question_key_term": "features",
      "term_explanation": "Whether the study utilizes additional derived features such as vegetation indices (NDVI, LAI, FAPAR), land surface temperature, albedo, emissivity, topographic information (e.g., DEM), or other external variables. These features can significantly enhance classification performance."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "2",
    "paper_title": "Mapping essential urban land use categories in China (EULUC-China): preliminary results for 2018",
    "question_id": "Q9",
    "question": "What type of model is implemented in this study?",
    "choices": {
      "A": "SVM",
      "B": "RF",
      "C": "XGBoost",
      "D": "CNN",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "B",
    "metadata": {
      "Task-oriented Category": "Technical approach & details",
      "question_key_term": "ML model",
      "term_explanation": "The classification model used to perform land cover or land use mapping. Options include traditional machine learning algorithms (e.g., random forest, SVM), deep learning architectures (e.g., CNNs, U-Net), or hybrid and transformer-based models. The model choice reflects the learning strategy and computational design."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "2",
    "paper_title": "Mapping essential urban land use categories in China (EULUC-China): preliminary results for 2018",
    "question_id": "Q10",
    "question": "What performance metrics are reported?",
    "choices": {
      "A": "Overall Accuracy (OA)",
      "B": "F1-score",
      "C": "Kappa",
      "D": "IoU / mIoU",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "A",
    "metadata": {
      "Task-oriented Category": "Technical approach & details",
      "question_key_term": "performance metrics",
      "term_explanation": "The quantitative indicators used to evaluate model performance, such as overall accuracy (OA), F1-score, Kappa coefficient, and mean intersection-over-union (mIoU). The selection of metrics affects the interpretability and comparability of results."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "2",
    "paper_title": "Mapping essential urban land use categories in China (EULUC-China): preliminary results for 2018",
    "question_id": "Q11",
    "question": "Is any comparative analysis included?",
    "choices": {
      "A": "Compared with traditional classifiers (e.g., RF, SVM)",
      "B": "Compared with deep models (e.g., U-Net variants)",
      "C": "Ablation study",
      "D": "No comparison reported",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "D",
    "metadata": {
      "Task-oriented Category": "Conclusions & results",
      "question_key_term": "comparative analysis",
      "term_explanation": "Whether the study conducts comparisons with baseline models, traditional classifiers, or variants of the proposed method (e.g., ablation study, benchmark comparison). This provides insight into model robustness and improvement over existing methods."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "2",
    "paper_title": "Mapping essential urban land use categories in China (EULUC-China): preliminary results for 2018",
    "question_id": "Q12",
    "question": "What is the reported overall accuracy (OA)?",
    "choices": {
      "A": "40.6%",
      "B": "57.5%",
      "C": "61.2%",
      "D": "64.1%",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "A",
    "metadata": {
      "Task-oriented Category": "Conclusions & results",
      "question_key_term": "overall accuracy",
      "term_explanation": "The percentage of correctly classified samples over the total number of samples, as reported in the study. OA is a commonly used metric for remote sensing classification and provides a general measure of model performance."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "3",
    "paper_title": "Finer resolution observation and monitoring of global land cover: first mapping results with Landsat TM and ETM+ data",
    "question_id": "Q1",
    "question": "What is the number of land-cover / land-use classes classified in this study?",
    "choices": {
      "A": "2",
      "B": "3",
      "C": "9",
      "D": "17",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "C",
    "metadata": {
      "Task-oriented Category": "Study subject & experimental setup",
      "question_key_term": "land cover",
      "term_explanation": "The total number of land cover or land use categories defined for classification. This reflects the complexity of the classification task and determines how fine-grained the land surface representation is. Typical classes include urban, vegetation, water, bare land, etc."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "3",
    "paper_title": "Finer resolution observation and monitoring of global land cover: first mapping results with Landsat TM and ETM+ data",
    "question_id": "Q2",
    "question": "What is the spatial extent of the study area?",
    "choices": {
      "A": "6,229 km²",
      "B": "100,000 km²",
      "C": "250,000 km²",
      "D": "656,889 km²",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "F",
    "metadata": {
      "Task-oriented Category": "Study subject & experimental setup",
      "question_key_term": "geographic area",
      "term_explanation": "The geographic area covered by the study, typically measured in square kilometers. This affects the generalizability and computational complexity of the study and can range from local sites to national or global scales."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "3",
    "paper_title": "Finer resolution observation and monitoring of global land cover: first mapping results with Landsat TM and ETM+ data",
    "question_id": "Q3",
    "question": "What is the geographic type of the study area?",
    "choices": {
      "A": "Urban",
      "B": "Rural",
      "C": "Mixed",
      "D": "Natural (e.g., forest, wetland, desert)",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "C",
    "metadata": {
      "Task-oriented Category": "Study subject & experimental setup",
      "question_key_term": "geographic type",
      "term_explanation": "The  geographic type or functional nature of the study region. Common categories include urban, rural, natural ecosystems (e.g., forest, wetland), or mixed types. The geographic context influences the spectral heterogeneity and classification difficulty."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "3",
    "paper_title": "Finer resolution observation and monitoring of global land cover: first mapping results with Landsat TM and ETM+ data",
    "question_id": "Q4",
    "question": "What is the temporal scope of the data used?",
    "choices": {
      "A": "Single-scene imagery",
      "B": "Short-term imagery ( ≤3 months)",
      "C": "Mid-term imagery ( >3 and ≤12 months)",
      "D": "Long-term imagery ( >1 year)",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "D",
    "metadata": {
      "Task-oriented Category": "Data characteristics & collection",
      "question_key_term": "time span",
      "term_explanation": "The total time span covered by the remote sensing imagery used in the study. This could be a single image, a short-term period (weeks to months), seasonal cycles, or long-term historical records (multi-year time series)."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "3",
    "paper_title": "Finer resolution observation and monitoring of global land cover: first mapping results with Landsat TM and ETM+ data",
    "question_id": "Q5",
    "question": "What type of remote sensing data is used?",
    "choices": {
      "A": "Optical",
      "B": "SAR",
      "C": "LiDAR",
      "D": "Multisource fusion",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "A",
    "metadata": {
      "Task-oriented Category": "Data characteristics & collection",
      "question_key_term": "data type",
      "term_explanation": "The primary data modality employed in the study. Categories include optical imagery, synthetic aperture radar (SAR), LiDAR, or combinations of multiple sources. The data type determines the spectral, spatial, and structural information available for classification."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "3",
    "paper_title": "Finer resolution observation and monitoring of global land cover: first mapping results with Landsat TM and ETM+ data",
    "question_id": "Q6",
    "question": "Which specific satellite data is used?",
    "choices": {
      "A": "Sentinel-1",
      "B": "Landsat series",
      "C": "Sentinel-2",
      "D": "PlanetScope",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "B",
    "metadata": {
      "Task-oriented Category": "Data characteristics & collection",
      "question_key_term": "satellite",
      "term_explanation": "The name(s) of satellite missions providing the source imagery, such as Sentinel-2, Landsat-8, Sentinel-1, MODIS, or commercial platforms like Planet or WorldView. This helps determine resolution, revisit frequency, and sensor characteristics."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "3",
    "paper_title": "Finer resolution observation and monitoring of global land cover: first mapping results with Landsat TM and ETM+ data",
    "question_id": "Q7",
    "question": "What is the spatial resolution of the primary imagery used?",
    "choices": {
      "A": "10 m",
      "B": "18 m",
      "C": "30 m",
      "D": "60 m",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "C",
    "metadata": {
      "Task-oriented Category": "Data characteristics & collection",
      "question_key_term": "spatial resolution",
      "term_explanation": "The ground sampling distance (in meters) of the input imagery, indicating the size of the smallest detectable object. High-resolution imagery (e.g., <5 m) allows fine-scale mapping, while moderate to coarse resolutions (e.g., 30 m or >100 m) are suited for regional to global scales."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "3",
    "paper_title": "Finer resolution observation and monitoring of global land cover: first mapping results with Landsat TM and ETM+ data",
    "question_id": "Q8",
    "question": "Are auxiliary features used beyond raw spectral bands?",
    "choices": {
      "A": "Vegetation indices (e.g., EVI)",
      "B": "Water features (e.g., NDWI)",
      "C": "Vegetation indices and Water indices",
      "D": "Elevation / DEM",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "A",
    "metadata": {
      "Task-oriented Category": "Data characteristics & collection",
      "question_key_term": "features",
      "term_explanation": "Whether the study utilizes additional derived features such as vegetation indices (NDVI, LAI, FAPAR), land surface temperature, albedo, emissivity, topographic information (e.g., DEM), or other external variables. These features can significantly enhance classification performance."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "3",
    "paper_title": "Finer resolution observation and monitoring of global land cover: first mapping results with Landsat TM and ETM+ data",
    "question_id": "Q9",
    "question": "What type of model is implemented in this study?",
    "choices": {
      "A": "SVM",
      "B": "RF",
      "C": "J4.8 Classifier",
      "D": "MLC",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "E",
    "metadata": {
      "Task-oriented Category": "Technical approach & details",
      "question_key_term": "ML model",
      "term_explanation": "The classification model used to perform land cover or land use mapping. Options include traditional machine learning algorithms (e.g., random forest, SVM), deep learning architectures (e.g., CNNs, U-Net), or hybrid and transformer-based models. The model choice reflects the learning strategy and computational design."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "3",
    "paper_title": "Finer resolution observation and monitoring of global land cover: first mapping results with Landsat TM and ETM+ data",
    "question_id": "Q10",
    "question": "What performance metrics are reported?",
    "choices": {
      "A": "Overall Accuracy (OA)",
      "B": "F1-score",
      "C": "Kappa",
      "D": "OA and Kappa",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "A",
    "metadata": {
      "Task-oriented Category": "Technical approach & details",
      "question_key_term": "performance metrics",
      "term_explanation": "The quantitative indicators used to evaluate model performance, such as overall accuracy (OA), F1-score, Kappa coefficient, and mean intersection-over-union (mIoU). The selection of metrics affects the interpretability and comparability of results."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "3",
    "paper_title": "Finer resolution observation and monitoring of global land cover: first mapping results with Landsat TM and ETM+ data",
    "question_id": "Q11",
    "question": "Is any comparative analysis included?",
    "choices": {
      "A": "Compared with traditional classifiers (e.g., RF, SVM)",
      "B": "Compared with deep models (e.g., U-Net variants)",
      "C": "Ablation study",
      "D": "No comparison reported",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "A",
    "metadata": {
      "Task-oriented Category": "Conclusions & results",
      "question_key_term": "comparative analysis",
      "term_explanation": "Whether the study conducts comparisons with baseline models, traditional classifiers, or variants of the proposed method (e.g., ablation study, benchmark comparison). This provides insight into model robustness and improvement over existing methods."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "3",
    "paper_title": "Finer resolution observation and monitoring of global land cover: first mapping results with Landsat TM and ETM+ data",
    "question_id": "Q12",
    "question": "What is the reported overall accuracy (OA)?",
    "choices": {
      "A": "53.88%",
      "B": "57.88%",
      "C": "59.83%",
      "D": "64.89%",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "D",
    "metadata": {
      "Task-oriented Category": "Conclusions & results",
      "question_key_term": "overall accuracy",
      "term_explanation": "The percentage of correctly classified samples over the total number of samples, as reported in the study. OA is a commonly used metric for remote sensing classification and provides a general measure of model performance."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "4",
    "paper_title": "Annual dynamics of global land cover and its long-term changes from 1982 to 2015",
    "question_id": "Q1",
    "question": "What is the number of land-cover / land-use classes classified in this study?",
    "choices": {
      "A": "1",
      "B": "7",
      "C": "11",
      "D": "20",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "B",
    "metadata": {
      "Task-oriented Category": "Study subject & experimental setup",
      "question_key_term": "land cover",
      "term_explanation": "The total number of land cover or land use categories defined for classification. This reflects the complexity of the classification task and determines how fine-grained the land surface representation is. Typical classes include urban, vegetation, water, bare land, etc."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "4",
    "paper_title": "Annual dynamics of global land cover and its long-term changes from 1982 to 2015",
    "question_id": "Q2",
    "question": "What is the spatial extent of the study area?",
    "choices": {
      "A": "67,000 km²",
      "B": "132,000 km²",
      "C": "151,942 km²",
      "D": "315,000 km²",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "F",
    "metadata": {
      "Task-oriented Category": "Study subject & experimental setup",
      "question_key_term": "geographic area",
      "term_explanation": "The geographic area covered by the study, typically measured in square kilometers. This affects the generalizability and computational complexity of the study and can range from local sites to national or global scales."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "4",
    "paper_title": "Annual dynamics of global land cover and its long-term changes from 1982 to 2015",
    "question_id": "Q3",
    "question": "What is the geographic type of the study area?",
    "choices": {
      "A": "Urban",
      "B": "Rural",
      "C": "Mixed",
      "D": "Natural (e.g., forest, wetland, desert)",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "C",
    "metadata": {
      "Task-oriented Category": "Study subject & experimental setup",
      "question_key_term": "geographic type",
      "term_explanation": "The  geographic type or functional nature of the study region. Common categories include urban, rural, natural ecosystems (e.g., forest, wetland), or mixed types. The geographic context influences the spectral heterogeneity and classification difficulty."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "4",
    "paper_title": "Annual dynamics of global land cover and its long-term changes from 1982 to 2015",
    "question_id": "Q4",
    "question": "What is the temporal scope of the data used?",
    "choices": {
      "A": "Single-scene imagery",
      "B": "Short-term imagery ( ≤3 months)",
      "C": "Mid-term imagery ( >3 and ≤12 months)",
      "D": "Long-term imagery ( >1 year)",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "D",
    "metadata": {
      "Task-oriented Category": "Data characteristics & collection",
      "question_key_term": "time span",
      "term_explanation": "The total time span covered by the remote sensing imagery used in the study. This could be a single image, a short-term period (weeks to months), seasonal cycles, or long-term historical records (multi-year time series)."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "4",
    "paper_title": "Annual dynamics of global land cover and its long-term changes from 1982 to 2015",
    "question_id": "Q5",
    "question": "What type of remote sensing data is used?",
    "choices": {
      "A": "Optical",
      "B": "SAR",
      "C": "LiDAR",
      "D": "Multisource fusion",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "D",
    "metadata": {
      "Task-oriented Category": "Data characteristics & collection",
      "question_key_term": "data type",
      "term_explanation": "The primary data modality employed in the study. Categories include optical imagery, synthetic aperture radar (SAR), LiDAR, or combinations of multiple sources. The data type determines the spectral, spatial, and structural information available for classification."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "4",
    "paper_title": "Annual dynamics of global land cover and its long-term changes from 1982 to 2015",
    "question_id": "Q6",
    "question": "Which specific satellite data is used?",
    "choices": {
      "A": "Sentinel-1",
      "B": "Sentinel-2",
      "C": "Luojia-1",
      "D": "Multisources",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "D",
    "metadata": {
      "Task-oriented Category": "Data characteristics & collection",
      "question_key_term": "satellite",
      "term_explanation": "The name(s) of satellite missions providing the source imagery, such as Sentinel-2, Landsat-8, Sentinel-1, MODIS, or commercial platforms like Planet or WorldView. This helps determine resolution, revisit frequency, and sensor characteristics."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "4",
    "paper_title": "Annual dynamics of global land cover and its long-term changes from 1982 to 2015",
    "question_id": "Q7",
    "question": "What is the spatial resolution of the primary imagery used?",
    "choices": {
      "A": "5 m",
      "B": "10 m",
      "C": "30 m",
      "D": "5 km",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "E",
    "metadata": {
      "Task-oriented Category": "Data characteristics & collection",
      "question_key_term": "spatial resolution",
      "term_explanation": "The ground sampling distance (in meters) of the input imagery, indicating the size of the smallest detectable object. High-resolution imagery (e.g., <5 m) allows fine-scale mapping, while moderate to coarse resolutions (e.g., 30 m or >100 m) are suited for regional to global scales."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "4",
    "paper_title": "Annual dynamics of global land cover and its long-term changes from 1982 to 2015",
    "question_id": "Q8",
    "question": "Are auxiliary features used beyond raw spectral bands?",
    "choices": {
      "A": "Vegetation indices only (e.g., NDVI, LAI, FAPAR)",
      "B": "Vegetation + energy fluxes (e.g., ET, GPP)",
      "C": "Vegetation + albedo/emissivity (e.g., BBE, white-sky albedo)",
      "D": "Albedo/emissivity",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "C",
    "metadata": {
      "Task-oriented Category": "Data characteristics & collection",
      "question_key_term": "features",
      "term_explanation": "Whether the study utilizes additional derived features such as vegetation indices (NDVI, LAI, FAPAR), land surface temperature, albedo, emissivity, topographic information (e.g., DEM), or other external variables. These features can significantly enhance classification performance."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "4",
    "paper_title": "Annual dynamics of global land cover and its long-term changes from 1982 to 2015",
    "question_id": "Q9",
    "question": "What type of model is implemented in this study?",
    "choices": {
      "A": "SVM",
      "B": "RF",
      "C": "XGBoost",
      "D": "CNN",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "B",
    "metadata": {
      "Task-oriented Category": "Technical approach & details",
      "question_key_term": "ML model",
      "term_explanation": "The classification model used to perform land cover or land use mapping. Options include traditional machine learning algorithms (e.g., random forest, SVM), deep learning architectures (e.g., CNNs, U-Net), or hybrid and transformer-based models. The model choice reflects the learning strategy and computational design."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "4",
    "paper_title": "Annual dynamics of global land cover and its long-term changes from 1982 to 2015",
    "question_id": "Q10",
    "question": "What performance metrics are reported?",
    "choices": {
      "A": "Overall Accuracy (OA)",
      "B": "F1-score",
      "C": "Kappa",
      "D": "OA and Kappa",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "A",
    "metadata": {
      "Task-oriented Category": "Technical approach & details",
      "question_key_term": "performance metrics",
      "term_explanation": "The quantitative indicators used to evaluate model performance, such as overall accuracy (OA), F1-score, Kappa coefficient, and mean intersection-over-union (mIoU). The selection of metrics affects the interpretability and comparability of results."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "4",
    "paper_title": "Annual dynamics of global land cover and its long-term changes from 1982 to 2015",
    "question_id": "Q11",
    "question": "Is any comparative analysis included?",
    "choices": {
      "A": "Compared with traditional classifiers (e.g., RF, SVM)",
      "B": "Compared with deep models (e.g., U-Net variants)",
      "C": "Compared with previous products",
      "D": "No comparison reported",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "C",
    "metadata": {
      "Task-oriented Category": "Conclusions & results",
      "question_key_term": "comparative analysis",
      "term_explanation": "Whether the study conducts comparisons with baseline models, traditional classifiers, or variants of the proposed method (e.g., ablation study, benchmark comparison). This provides insight into model robustness and improvement over existing methods."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "4",
    "paper_title": "Annual dynamics of global land cover and its long-term changes from 1982 to 2015",
    "question_id": "Q12",
    "question": "What is the reported overall accuracy (OA)?",
    "choices": {
      "A": "73.54%",
      "B": "86.51%",
      "C": "87.12%",
      "D": "92.26%",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "B",
    "metadata": {
      "Task-oriented Category": "Conclusions & results",
      "question_key_term": "overall accuracy",
      "term_explanation": "The percentage of correctly classified samples over the total number of samples, as reported in the study. OA is a commonly used metric for remote sensing classification and provides a general measure of model performance."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "5",
    "paper_title": "Annual maps of global artificial impervious area (GAIA) between 1985 and 2018",
    "question_id": "Q1",
    "question": "What is the number of land-cover / land-use classes classified in this study?",
    "choices": {
      "A": "1",
      "B": "3",
      "C": "34",
      "D": "155",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "A",
    "metadata": {
      "Task-oriented Category": "Study subject & experimental setup",
      "question_key_term": "land cover",
      "term_explanation": "The total number of land cover or land use categories defined for classification. This reflects the complexity of the classification task and determines how fine-grained the land surface representation is. Typical classes include urban, vegetation, water, bare land, etc."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "5",
    "paper_title": "Annual maps of global artificial impervious area (GAIA) between 1985 and 2018",
    "question_id": "Q2",
    "question": "What is the spatial extent of the study area?",
    "choices": {
      "A": "108,962 km²",
      "B": "340,625 km²",
      "C": "218,859 km²",
      "D": "797,076 km²",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "D",
    "metadata": {
      "Task-oriented Category": "Study subject & experimental setup",
      "question_key_term": "geographic area",
      "term_explanation": "The geographic area covered by the study, typically measured in square kilometers. This affects the generalizability and computational complexity of the study and can range from local sites to national or global scales."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "5",
    "paper_title": "Annual maps of global artificial impervious area (GAIA) between 1985 and 2018",
    "question_id": "Q3",
    "question": "What is the geographic type of the study area?",
    "choices": {
      "A": "Urban",
      "B": "Rural",
      "C": "Mixed",
      "D": "Natural (e.g., forest, wetland, desert)",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "A",
    "metadata": {
      "Task-oriented Category": "Study subject & experimental setup",
      "question_key_term": "geographic type",
      "term_explanation": "The  geographic type or functional nature of the study region. Common categories include urban, rural, natural ecosystems (e.g., forest, wetland), or mixed types. The geographic context influences the spectral heterogeneity and classification difficulty."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "5",
    "paper_title": "Annual maps of global artificial impervious area (GAIA) between 1985 and 2018",
    "question_id": "Q4",
    "question": "What is the temporal scope of the data used?",
    "choices": {
      "A": "Single-scene imagery",
      "B": "Short-term imagery ( ≤3 months)",
      "C": "Mid-term imagery ( >3 and ≤12 months)",
      "D": "Long-term imagery ( >1 year)",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "D",
    "metadata": {
      "Task-oriented Category": "Data characteristics & collection",
      "question_key_term": "time span",
      "term_explanation": "The total time span covered by the remote sensing imagery used in the study. This could be a single image, a short-term period (weeks to months), seasonal cycles, or long-term historical records (multi-year time series)."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "5",
    "paper_title": "Annual maps of global artificial impervious area (GAIA) between 1985 and 2018",
    "question_id": "Q5",
    "question": "What type of remote sensing data is used?",
    "choices": {
      "A": "Optical",
      "B": "SAR",
      "C": "LiDAR",
      "D": "Multisource fusion",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "D",
    "metadata": {
      "Task-oriented Category": "Data characteristics & collection",
      "question_key_term": "data type",
      "term_explanation": "The primary data modality employed in the study. Categories include optical imagery, synthetic aperture radar (SAR), LiDAR, or combinations of multiple sources. The data type determines the spectral, spatial, and structural information available for classification."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "5",
    "paper_title": "Annual maps of global artificial impervious area (GAIA) between 1985 and 2018",
    "question_id": "Q6",
    "question": "Which specific satellite data is used?",
    "choices": {
      "A": "Sentinel-1",
      "B": "Landsat series",
      "C": "VIIRS NTL",
      "D": "Landsat series, Sentinel-1 and VIIRS NTL",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "D",
    "metadata": {
      "Task-oriented Category": "Data characteristics & collection",
      "question_key_term": "satellite",
      "term_explanation": "The name(s) of satellite missions providing the source imagery, such as Sentinel-2, Landsat-8, Sentinel-1, MODIS, or commercial platforms like Planet or WorldView. This helps determine resolution, revisit frequency, and sensor characteristics."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "5",
    "paper_title": "Annual maps of global artificial impervious area (GAIA) between 1985 and 2018",
    "question_id": "Q7",
    "question": "What is the spatial resolution of the primary imagery used?",
    "choices": {
      "A": "10 m",
      "B": "30 m",
      "C": "100 m",
      "D": "250 m",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "B",
    "metadata": {
      "Task-oriented Category": "Data characteristics & collection",
      "question_key_term": "spatial resolution",
      "term_explanation": "The ground sampling distance (in meters) of the input imagery, indicating the size of the smallest detectable object. High-resolution imagery (e.g., <5 m) allows fine-scale mapping, while moderate to coarse resolutions (e.g., 30 m or >100 m) are suited for regional to global scales."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "5",
    "paper_title": "Annual maps of global artificial impervious area (GAIA) between 1985 and 2018",
    "question_id": "Q8",
    "question": "Are auxiliary features used beyond raw spectral bands?",
    "choices": {
      "A": "Vegetation indices (e.g., EVI)",
      "B": "Vegetation + energy fluxes (e.g., ET, GPP)",
      "C": "Water features (e.g., NDWI, MNDWI)",
      "D": "Vegetation indices and Water indices",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "C",
    "metadata": {
      "Task-oriented Category": "Data characteristics & collection",
      "question_key_term": "features",
      "term_explanation": "Whether the study utilizes additional derived features such as vegetation indices (NDVI, LAI, FAPAR), land surface temperature, albedo, emissivity, topographic information (e.g., DEM), or other external variables. These features can significantly enhance classification performance."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "5",
    "paper_title": "Annual maps of global artificial impervious area (GAIA) between 1985 and 2018",
    "question_id": "Q9",
    "question": "What type of model is implemented in this study?",
    "choices": {
      "A": "Spatially Explicit",
      "B": "Temporal Consistency",
      "C": "Spatially Explicit and Temporal Consistency",
      "D": "Transformer",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "C",
    "metadata": {
      "Task-oriented Category": "Technical approach & details",
      "question_key_term": "ML model",
      "term_explanation": "The classification model used to perform land cover or land use mapping. Options include traditional machine learning algorithms (e.g., random forest, SVM), deep learning architectures (e.g., CNNs, U-Net), or hybrid and transformer-based models. The model choice reflects the learning strategy and computational design."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "5",
    "paper_title": "Annual maps of global artificial impervious area (GAIA) between 1985 and 2018",
    "question_id": "Q10",
    "question": "What performance metrics are reported?",
    "choices": {
      "A": "Overall Accuracy (OA)",
      "B": "F1-score",
      "C": "Kappa",
      "D": "OA and Kappa",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "A",
    "metadata": {
      "Task-oriented Category": "Technical approach & details",
      "question_key_term": "performance metrics",
      "term_explanation": "The quantitative indicators used to evaluate model performance, such as overall accuracy (OA), F1-score, Kappa coefficient, and mean intersection-over-union (mIoU). The selection of metrics affects the interpretability and comparability of results."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "5",
    "paper_title": "Annual maps of global artificial impervious area (GAIA) between 1985 and 2018",
    "question_id": "Q11",
    "question": "Is any comparative analysis included?",
    "choices": {
      "A": "Compared with traditional classifiers (e.g., RF, SVM)",
      "B": "Compared with deep models (e.g., U-Net variants)",
      "C": "Compared with previous products",
      "D": "No comparison reported",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "C",
    "metadata": {
      "Task-oriented Category": "Conclusions & results",
      "question_key_term": "comparative analysis",
      "term_explanation": "Whether the study conducts comparisons with baseline models, traditional classifiers, or variants of the proposed method (e.g., ablation study, benchmark comparison). This provides insight into model robustness and improvement over existing methods."
    }
  },
  {
    "subject": "Earth Science - Remote Sensing",
    "paper_id": "5",
    "paper_title": "Annual maps of global artificial impervious area (GAIA) between 1985 and 2018",
    "question_id": "Q12",
    "question": "What is the reported overall accuracy (OA)?",
    "choices": {
      "A": "15%",
      "B": "43%",
      "C": "70%",
      "D": "89%",
      "E": "All of above",
      "F": "None of above"
    },
    "answer": "D",
    "metadata": {
      "Task-oriented Category": "Conclusions & results",
      "question_key_term": "overall accuracy",
      "term_explanation": "The percentage of correctly classified samples over the total number of samples, as reported in the study. OA is a commonly used metric for remote sensing classification and provides a general measure of model performance."
    }
  }
]