grounded_scan

  • توضیحات :

Grounded SCAN (gSCAN) یک مجموعه داده مصنوعی برای ارزیابی تعمیم ترکیبی در درک زبان موقعیتی است. gSCAN دستورالعمل‌های زبان طبیعی را با توالی‌های عمل جفت می‌کند و از عامل می‌خواهد دستورالعمل‌ها را در چارچوب یک محیط ناوبری بصری مبتنی بر شبکه تفسیر کند.

اطلاعات بیشتر را می توان در آدرس زیر یافت:

FeaturesDict({
    'command': Sequence(Text(shape=(), dtype=string)),
    'manner': Text(shape=(), dtype=string),
    'meaning': Sequence(Text(shape=(), dtype=string)),
    'referred_target': Text(shape=(), dtype=string),
    'situation': FeaturesDict({
        'agent_direction': int32,
        'agent_position': FeaturesDict({
            'column': int32,
            'row': int32,
        }),
        'direction_to_target': Text(shape=(), dtype=string),
        'distance_to_target': int32,
        'grid_size': int32,
        'placed_objects': Sequence({
            'object': FeaturesDict({
                'color': Text(shape=(), dtype=string),
                'shape': Text(shape=(), dtype=string),
                'size': int32,
            }),
            'position': FeaturesDict({
                'column': int32,
                'row': int32,
            }),
            'vector': Text(shape=(), dtype=string),
        }),
        'target_object': FeaturesDict({
            'object': FeaturesDict({
                'color': Text(shape=(), dtype=string),
                'shape': Text(shape=(), dtype=string),
                'size': int32,
            }),
            'position': FeaturesDict({
                'column': int32,
                'row': int32,
            }),
            'vector': Text(shape=(), dtype=string),
        }),
    }),
    'target_commands': Sequence(Text(shape=(), dtype=string)),
    'verb_in_command': Text(shape=(), dtype=string),
})
  • مستندات ویژگی :
ویژگی کلاس شکل نوع D شرح
FeaturesDict
فرمان دنباله (متن) (هیچ یک،) رشته
شیوه متن رشته
معنی دنباله (متن) (هیچ یک،) رشته
ارجاع_هدف متن رشته
وضعیت FeaturesDict
موقعیت/جهت_عامل تانسور int32
موقعیت/موقعیت_عامل FeaturesDict
وضعیت/موقعیت_عامل/ستون تانسور int32
وضعیت/موقعیت_عامل/ردیف تانسور int32
موقعیت/جهت_به_هدف متن رشته
موقعیت/فاصله_تا_هدف تانسور int32
وضعیت/شبکه_اندازه تانسور int32
موقعیت/اشیاء_قرار داده شده توالی
وضعیت/اشیاء_قرارداده/اشیاء FeaturesDict
وضعیت/اشیاء_قرارداده/اشیاء/رنگ متن رشته
وضعیت/اشیاء_قرار داده شده/اشیاء/شکل متن رشته
وضعیت/اشیاء_قرارداده/اشیاء/اندازه تانسور int32
وضعیت/اشیاء_قرار داده شده/موقعیت FeaturesDict
وضعیت/اشیاء_قرار داده شده/موقعیت/ستون تانسور int32
وضعیت/اشیاء_قرار داده شده/موقعیت/ردیف تانسور int32
وضعیت/اشیاء_قرار داده شده/بردار متن رشته
موقعیت/هدف_ابژه FeaturesDict
وضعیت/هدف_شیء/شیء FeaturesDict
status/target_object/object/color متن رشته
status/target_object/object/shape متن رشته
status/target_object/object/size تانسور int32
وضعیت/هدف_شیء/موقعیت FeaturesDict
status/target_object/position/column تانسور int32
status/target_object/position/row تانسور int32
وضعیت/هدف_شیء/بردار متن رشته
target_commands دنباله (متن) (هیچ یک،) رشته
فعل_در_فرمان متن رشته
@inproceedings{NEURIPS2020_e5a90182,
 author = {Ruis, Laura and Andreas, Jacob and Baroni, Marco and Bouchacourt, Diane and Lake, Brenden M},
 booktitle = {Advances in Neural Information Processing Systems},
 editor = {H. Larochelle and M. Ranzato and R. Hadsell and M. F. Balcan and H. Lin},
 pages = {19861--19872},
 publisher = {Curran Associates, Inc.},
 title = {A Benchmark for Systematic Generalization in Grounded Language Understanding},
 url = {https://proceedings.neurips.cc/paper/2020/file/e5a90182cc81e12ab5e72d66e0b46fe3-Paper.pdf},
 volume = {33},
 year = {2020}
}

@inproceedings{qiu-etal-2021-systematic,
    title = "Systematic Generalization on g{SCAN}: {W}hat is Nearly Solved and What is Next?",
    author = "Qiu, Linlu  and
      Hu, Hexiang  and
      Zhang, Bowen  and
      Shaw, Peter  and
      Sha, Fei",
    booktitle = "Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing",
    month = nov,
    year = "2021",
    address = "Online and Punta Cana, Dominican Republic",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2021.emnlp-main.166",
    doi = "10.18653/v1/2021.emnlp-main.166",
    pages = "2180--2188",
}

grounded_scan/compositional_splits (پیکربندی پیش‌فرض)

  • توضیحات پیکربندی : نمونه هایی برای تعمیم ترکیبی.

  • حجم دانلود : 82.10 MiB

  • حجم مجموعه داده : 998.11 MiB

  • تقسیم ها :

شکاف مثال ها
'adverb_1' 112880
'adverb_2' 38582
'contextual' 11,460
'dev' 3716
'situational_1' 88642
'situational_2' 16808
'test' 19,282
'train' 367,933
'visual' 37,436
'visual_easier' 18718

grounded_scan/target_length_split

  • توضیحات پیکربندی : نمونه هایی برای تعمیم به طول های هدف بزرگتر.

  • حجم دانلود : 53.41 MiB

  • حجم مجموعه داده : 546.73 MiB

  • تقسیم ها :

شکاف مثال ها
'dev' 1,821
'target_lengths' 198588
'test' 37784
'train' 180,301

grounded_scan/spatial_relation splits

  • توضیحات پیکربندی : نمونه هایی برای استدلال رابطه فضایی.

  • حجم دانلود : 89.59 MiB

  • حجم مجموعه داده : 675.09 MiB

  • تقسیم ها :

شکاف مثال ها
'dev' 2,617
'referent' 30,492
'relation' 6285
'relative_position_1' 41576
'relative_position_2' 41,529
'test' 28,526
'train' 259,088
'visual' 62250