Bangla News Headlines with Paired Images for Multimodal Classification

20,000 Bangla news headlines are paired with corresponding images for multimodal classification tasks. The dataset is hosted on Kaggle, but details about its author, organization, and creation date are unknown. Column-level documentation and file formats are also unspecified.

Use Cases

• Train multimodal classification models based on paired headlines and images.
• Benchmark Bangla language understanding models using news content.
• Develop image-text retrieval systems for Bangla news articles.
• Study cross-modal representation learning for low-resource languages.

Strengths

• Contains 20,000 data points, providing a substantial sample for model training.
• Explicitly designed for multimodal tasks, pairing text headlines with images.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is known but other scale details like file size and formats are unknown.
• Data may reflect geographic and source bias inherent to its collection from Kaggle.

Provenance

Source

Kaggle

Geography

Bangla language region (likely Bangladesh or West Bengal, India)