Dataset Preview
Duplicate
The full dataset viewer is not available (click to read why). Only showing a preview of the rows.
The dataset generation failed because of a cast error
Error code:   DatasetGenerationCastError
Exception:    DatasetGenerationCastError
Message:      An error occurred while generating the dataset

All the data files must have the same columns, but at some point there are 4 new columns ({'scene_description', 'ocr', 'culture_comment', 'problem_type'})

This happened while the csv dataset builder was generating data using

hf://datasets/SoulQrat/russian_memes/dataset_extended.csv (at revision fade22fbc64e0dccdd7d5e86729124c32f2037e9), [/tmp/hf-datasets-cache/medium/datasets/43420239151497-config-parquet-and-info-SoulQrat-russian_memes-ef77666e/hub/datasets--SoulQrat--russian_memes/snapshots/fade22fbc64e0dccdd7d5e86729124c32f2037e9/dataset_basic.csv (origin=hf://datasets/SoulQrat/russian_memes@fade22fbc64e0dccdd7d5e86729124c32f2037e9/dataset_basic.csv), /tmp/hf-datasets-cache/medium/datasets/43420239151497-config-parquet-and-info-SoulQrat-russian_memes-ef77666e/hub/datasets--SoulQrat--russian_memes/snapshots/fade22fbc64e0dccdd7d5e86729124c32f2037e9/dataset_extended.csv (origin=hf://datasets/SoulQrat/russian_memes@fade22fbc64e0dccdd7d5e86729124c32f2037e9/dataset_extended.csv), /tmp/hf-datasets-cache/medium/datasets/43420239151497-config-parquet-and-info-SoulQrat-russian_memes-ef77666e/hub/datasets--SoulQrat--russian_memes/snapshots/fade22fbc64e0dccdd7d5e86729124c32f2037e9/images.zip (origin=hf://datasets/SoulQrat/russian_memes@fade22fbc64e0dccdd7d5e86729124c32f2037e9/images.zip)]

Please either edit the data files to have matching columns, or separate them into different configurations (see docs at https://hf.co/docs/hub/datasets-manual-configuration#multiple-configurations)
Traceback:    Traceback (most recent call last):
                File "/usr/local/lib/python3.12/site-packages/datasets/builder.py", line 1800, in _prepare_split_single
                  writer.write_table(table)
                File "/usr/local/lib/python3.12/site-packages/datasets/arrow_writer.py", line 765, in write_table
                  self._write_table(pa_table, writer_batch_size=writer_batch_size)
                File "/usr/local/lib/python3.12/site-packages/datasets/arrow_writer.py", line 773, in _write_table
                  pa_table = table_cast(pa_table, self._schema)
                             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/table.py", line 2321, in table_cast
                  return cast_table_to_schema(table, schema)
                         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/table.py", line 2249, in cast_table_to_schema
                  raise CastError(
              datasets.table.CastError: Couldn't cast
              filename: string
              ocr: string
              scene_description: string
              toxic_label: int64
              culture_label: int64
              culture_comment: string
              sarcasm_label: int64
              irony_label: int64
              problem_type: int64
              -- schema metadata --
              pandas: '{"index_columns": [{"kind": "range", "name": null, "start": 0, "' + 1352
              to
              {'filename': Value('string'), 'toxic_label': Value('int64'), 'culture_label': Value('int64'), 'sarcasm_label': Value('int64'), 'irony_label': Value('int64')}
              because column names don't match
              
              During handling of the above exception, another exception occurred:
              
              Traceback (most recent call last):
                File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 1347, in compute_config_parquet_and_info_response
                  parquet_operations = convert_to_parquet(builder)
                                       ^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 980, in convert_to_parquet
                  builder.download_and_prepare(
                File "/usr/local/lib/python3.12/site-packages/datasets/builder.py", line 882, in download_and_prepare
                  self._download_and_prepare(
                File "/usr/local/lib/python3.12/site-packages/datasets/builder.py", line 943, in _download_and_prepare
                  self._prepare_split(split_generator, **prepare_split_kwargs)
                File "/usr/local/lib/python3.12/site-packages/datasets/builder.py", line 1646, in _prepare_split
                  for job_id, done, content in self._prepare_split_single(
                                               ^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/builder.py", line 1802, in _prepare_split_single
                  raise DatasetGenerationCastError.from_cast_error(
              datasets.exceptions.DatasetGenerationCastError: An error occurred while generating the dataset
              
              All the data files must have the same columns, but at some point there are 4 new columns ({'scene_description', 'ocr', 'culture_comment', 'problem_type'})
              
              This happened while the csv dataset builder was generating data using
              
              hf://datasets/SoulQrat/russian_memes/dataset_extended.csv (at revision fade22fbc64e0dccdd7d5e86729124c32f2037e9), [/tmp/hf-datasets-cache/medium/datasets/43420239151497-config-parquet-and-info-SoulQrat-russian_memes-ef77666e/hub/datasets--SoulQrat--russian_memes/snapshots/fade22fbc64e0dccdd7d5e86729124c32f2037e9/dataset_basic.csv (origin=hf://datasets/SoulQrat/russian_memes@fade22fbc64e0dccdd7d5e86729124c32f2037e9/dataset_basic.csv), /tmp/hf-datasets-cache/medium/datasets/43420239151497-config-parquet-and-info-SoulQrat-russian_memes-ef77666e/hub/datasets--SoulQrat--russian_memes/snapshots/fade22fbc64e0dccdd7d5e86729124c32f2037e9/dataset_extended.csv (origin=hf://datasets/SoulQrat/russian_memes@fade22fbc64e0dccdd7d5e86729124c32f2037e9/dataset_extended.csv), /tmp/hf-datasets-cache/medium/datasets/43420239151497-config-parquet-and-info-SoulQrat-russian_memes-ef77666e/hub/datasets--SoulQrat--russian_memes/snapshots/fade22fbc64e0dccdd7d5e86729124c32f2037e9/images.zip (origin=hf://datasets/SoulQrat/russian_memes@fade22fbc64e0dccdd7d5e86729124c32f2037e9/images.zip)]
              
              Please either edit the data files to have matching columns, or separate them into different configurations (see docs at https://hf.co/docs/hub/datasets-manual-configuration#multiple-configurations)

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

filename
string
toxic_label
int64
culture_label
int64
sarcasm_label
int64
irony_label
int64
img_0000.jpg
0
0
0
0
img_0001.jpg
0
0
0
0
img_0002.jpg
0
0
0
1
img_0003.jpg
0
0
0
0
img_0004.jpg
0
0
0
0
img_0005.jpg
0
0
0
1
img_0006.jpg
0
0
1
0
img_0007.jpg
0
1
0
1
img_0008.jpg
0
0
0
1
img_0009.jpg
0
0
1
0
img_0010.jpg
0
1
0
0
img_0011.jpg
0
0
0
1
img_0012.jpg
0
0
0
0
img_0013.jpg
0
0
0
0
img_0014.jpg
0
0
0
0
img_0015.jpg
0
0
0
1
img_0016.jpg
1
1
1
0
img_0017.jpg
0
0
0
0
img_0018.jpg
0
0
0
0
img_0019.jpg
0
0
0
1
img_0020.jpg
0
0
0
1
img_0021.jpg
0
1
0
0
img_0022.jpg
0
0
0
0
img_0023.jpg
0
0
1
0
img_0024.jpg
0
0
0
1
img_0025.jpg
0
1
1
0
img_0026.jpg
0
0
0
0
img_0027.jpg
0
0
0
0
img_0028.jpg
0
0
0
0
img_0029.jpg
0
0
0
0
img_0030.jpg
0
0
0
0
img_0031.jpg
0
0
0
0
img_0032.jpg
0
0
0
0
img_0033.jpg
0
0
0
1
img_0034.jpg
0
1
0
1
img_0035.jpg
0
0
0
1
img_0036.jpg
0
0
0
0
img_0037.jpg
0
0
0
0
img_0038.jpg
0
0
1
0
img_0039.jpg
0
0
0
0
img_0040.jpg
0
0
0
0
img_0041.jpg
0
1
0
0
img_0042.jpg
0
1
1
0
img_0043.jpg
0
1
0
0
img_0044.jpg
1
0
0
1
img_0045.jpg
1
1
0
1
img_0046.jpg
0
0
0
1
img_0047.jpg
0
0
0
0
img_0048.jpg
0
0
0
0
img_0049.jpg
0
0
0
0
img_0050.jpg
0
0
0
0
img_0051.jpg
0
0
0
1
img_0052.jpg
0
0
0
0
img_0053.jpg
0
0
0
1
img_0054.jpg
0
0
0
0
img_0055.jpg
0
0
0
0
img_0056.jpg
0
1
0
0
img_0057.jpg
0
1
0
0
img_0058.jpg
0
0
0
0
img_0059.jpg
0
1
0
1
img_0060.jpg
0
0
0
1
img_0061.jpg
0
0
0
1
img_0062.jpg
0
1
0
0
img_0063.jpg
0
0
0
0
img_0064.jpg
1
1
0
0
img_0065.jpg
0
1
1
0
img_0066.jpg
0
0
0
1
img_0067.jpg
0
0
0
0
img_0068.jpg
0
0
1
0
img_0069.jpg
0
0
0
1
img_0070.jpg
0
0
0
1
img_0071.jpg
0
0
1
0
img_0072.jpg
0
0
1
0
img_0073.jpg
0
0
1
0
img_0074.jpg
0
0
1
0
img_0075.jpg
0
0
0
0
img_0076.jpg
0
0
1
0
img_0077.jpg
0
0
0
1
img_0078.jpg
0
0
1
0
img_0079.jpg
0
0
0
0
img_0080.jpg
1
0
1
0
img_0081.jpg
0
0
0
1
img_0082.jpg
0
0
0
1
img_0083.jpg
0
0
1
0
img_0084.jpg
0
0
0
0
img_0085.jpg
0
0
1
0
img_0086.jpg
0
0
0
1
img_0087.jpg
0
0
0
1
img_0088.jpg
0
0
0
1
img_0089.jpg
0
0
0
1
img_0090.jpg
0
1
0
1
img_0091.jpg
0
1
0
0
img_0092.jpg
1
0
0
1
img_0093.jpg
0
1
0
0
img_0094.jpg
0
0
0
0
img_0095.jpg
0
0
0
0
img_0096.jpg
0
0
1
0
img_0097.jpg
0
0
0
0
img_0098.jpg
0
1
0
0
img_0099.jpg
0
1
0
0
End of preview.

Russian Meme Dataset

This dataset contains annotated Russian-language image memes. It is designed for evaluating multimodal and text-based models on meme understanding, with a focus on toxicity, sarcasm, irony, and local cultural references.

The dataset is released in two CSV versions:

  • dataset_basic.csv — a compact annotation table with image filenames and target labels.
  • dataset_extended.csv — an extended table that also includes OCR text, visual scene descriptions, and cultural-reference comments.

Files

dataset_basic.csv

This version is intended for standard classification experiments.

Column Type Description
filename string Image filename, for example img_0000.jpg.
toxic_label int, 0/1 Whether the meme contains content that should be treated as toxic or policy-violating.
culture_label int, 0/1 Whether the meme contains a Russian cultural reference, local meme format, or context-specific reference.
sarcasm_label int, 0/1 Whether the meme contains sarcasm.
irony_label int, 0/1 Whether the meme contains irony or post-ironic humor.

dataset_extended.csv

This version is intended for text-only experiments, error analysis, and interpretability studies.

Column Type Description
filename string Image filename, for example img_0000.jpg.
ocr string Text recognized directly from the meme image.
scene_description string Short textual description of the visual scene or objects in the meme.
toxic_label int, 0/1 Whether the meme contains content that should be treated as toxic or policy-violating.
culture_label int, 0/1 Whether the meme contains a Russian cultural reference, local meme format, or context-specific reference.
culture_comment string Explanation of the cultural reference, filled when culture_label = 1.
sarcasm_label int, 0/1 Whether the meme contains sarcasm.
irony_label int, 0/1 Whether the meme contains irony or post-ironic humor.
problem_type int, 0–3 Type of potentially difficult humor pattern.

Label definitions

Toxicity

toxic_label is assigned independently of the humor type.

  • 1: the meme contains or promotes policy-violating content, such as hate speech, harassment, dangerous content, graphic violence, extremist support, illegal goods, sexual content, self-harm encouragement, or other harmful categories.
  • 0: the meme does not fall into these categories. Non-toxic cases may still contain self-directed humor, absurd humor, political or historical satire without hate or harassment, profanity without a harmful target, or harmless wordplay.

A meme can be labeled as toxic even when the harmful content is presented as a joke, irony, dark humor, or absurdity.

Cultural reference

culture_label indicates whether the meme requires knowledge of Russian cultural context, local internet memes, social references, linguistic codes, or culturally specific background knowledge.

If culture_label = 1, the extended version provides culture_comment, which explains the reference.

Humor type

The humor labels describe the dominant rhetorical form of the meme.

  • sarcasm_label = 1: the meme uses sharp, biting, or mocking sarcasm.
  • irony_label = 1: the meme uses irony, post-irony, absurd contrast, or deliberate mismatch between literal and intended meaning.

The technical helper label other_label is not included in the released files. In the original annotation table, it was used only to ensure that sarcasm, irony, and other humor forms were mutually exclusive.

Difficult humor pattern

problem_type marks meme types that may be especially difficult for automated models. It is not a toxicity label.

Value Meaning Description
0 None The meme does not belong to one of the difficult humor categories below.
1 Self-directed or depressive humor The meme jokes about the author, personal failure, appearance, mental state, or similar self-directed topics.
2 Non-targeted aggression The meme contains profanity, aggression, or insults aimed at an abstract situation, object, or general frustration rather than a specific person or protected group.
3 Absurd or chaotic humor The meme mentions strange, frightening, or formally problematic things, but the surrounding absurdity makes the unrealistic and comic framing clear.
Downloads last month
85