grounded_scan

• Deskripsi :

Grounded SCAN (gSCAN) adalah kumpulan data sintetik untuk mengevaluasi generalisasi komposisi dalam pemahaman bahasa situasional. gSCAN memasangkan instruksi bahasa alami dengan urutan tindakan, dan mengharuskan agen untuk menginterpretasikan instruksi dalam konteks lingkungan navigasi visual berbasis grid.

Informasi lebih lanjut dapat ditemukan di:

• Untuk compositional_splits dan target_length_split : https://github.com/LauraRuis/groundedSCAN

• Untuk spatial_relation_splits : https://github.com/google-research/language/tree/master/language/gscan/data

• Beranda : https://github.com/LauraRuis/groundedSCAN

• Kode sumber : tfds.vision_language.grounded_scan.GroundedScan

• Versi :

  • 1.0.0 : Rilis awal.
  • 1.1.0 : Mengubah fitur vector menjadi Teks().
  • 2.0.0 (default): Menambahkan konfigurasi spatial_relation_splits baru.

• Di-cache otomatis ( dokumentasi ): Tidak

• Struktur fitur :

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),
})

• Dokumentasi fitur :

• Kunci yang diawasi (Lihat as_supervised doc ): None

• Gambar ( tfds.show_examples ): Tidak didukung.

• Kutipan :

@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 (konfigurasi default)

• Deskripsi konfigurasi : Contoh untuk generalisasi komposisi.

• Ukuran unduhan : 82.10 MiB

• Ukuran dataset : 998.11 MiB

• Perpecahan :

• Contoh ( tfds.as_dataframe ):

grounded_scan/target_length_split

• Deskripsi konfigurasi : Contoh untuk menggeneralisasi ke panjang target yang lebih besar.

• Ukuran unduhan : 53.41 MiB

• Ukuran dataset : 546.73 MiB

• Perpecahan :

• Contoh ( tfds.as_dataframe ):

grounded_scan/spatial_relation_splits

• Deskripsi konfigurasi : Contoh untuk penalaran hubungan spasial.

• Ukuran unduhan : 89.59 MiB

• Ukuran dataset : 675.09 MiB

• Perpecahan :

• Contoh ( tfds.as_dataframe ):