VALUABLY-net/iab-taxonomy-multilang-content

IAB Taxonomy Multilingual Content Dataset

Dataset Description

This dataset contains multilingual text data for IAB (Interactive Advertising Bureau) taxonomy classification.

The data files (train.csv, val.csv) were generated by processing, cleaning, and restructuring data file from the IAB-Taxonomy URL Content Dataset originally published on Kaggle.

The accompanying processing scripts and JSON mapping files are provided to support the data preparation and model training workflow.

Source Data & Attribution

This dataset is a derivative work based on the following:

• —Original Dataset: IAB-Taxonomy URL Content Dataset
• —Original Author: Ishikaaaaaathakur on Kaggle

License

The data files (train.csv, val.csv) are derivative works licensed under the Creative Commons Attribution-ShareAlike 4.0 International License (CC BY-SA 4.0), in accordance with the original dataset's license.

Dataset Structure

Data Files

• —train.csv: Training data with text content and IAB labels.
• —val.csv: Validation data.

Reference Files

• —iab_taxonomy.json: Complete IAB Taxonomy v2 reference.
• —dataset_to_iab_mapping.json: Mapping from original dataset categories to IAB codes.
• —label_map.json: Label mappings for model training.
• —id_to_label_map.json: ID to label mappings.
• —code_to_label_map.json: IAB code to label mappings.
• —label_to_code_map.json: Label to IAB code mappings.

Languages Supported

• —English
• —German
• —French

Usage

To load this dataset from Hugging Face:

from datasets import load_dataset

dataset = load_dataset("VALUABLY-net/iab-taxonomy-multilang-content")
print(dataset)

Label Information

The dataset uses IAB Tier 2 categories (392 labels total). Each sample is assigned a single IAB code for multi-class classification.