Multimodal Sarcasm Detection Dataset

Multimodal Sarcasm Detection is a dataset for detecting sarcasm from multiple data modalities, likely combining text and visual information. The dataset was created by author Carol99 and was last updated on Hugging Face in April 2022. Specific details on the number of samples, features, and collection method are not provided in the available metadata.

Use Cases

• Train a multimodal classifier to predict sarcasm labels from paired text captions and associated images.
• Fine-tune a vision-language model using visual features and textual context to recognize ironic intent.
• Benchmark model performance on sarcasm detection by analyzing predictions against annotated ground truth labels.
• Analyze the correlation between specific visual cues, such as facial expressions in images, and sarcastic textual content.

Strengths

• Dataset is hosted on the Hugging Face platform, facilitating access for the ML community.
• Focuses on the complex task of multimodal sarcasm detection, a niche research area.

Limitations

• No metadata on sample size, column names, or data balance is available, hindering reproducibility.
• The dataset's last known update was in April 2022, which may limit its relevance for current model training.

Provenance

Source

Hugging Face, uploaded by user Carol99.

Freshness

Last updated on 2022-04-25.