CLICK-ID: A novel dataset for Indonesian clickbait headlines

Area of Science:

• Natural Language Processing (NLP)
• Computational Linguistics
• Information Retrieval

Background:

• Clickbait detection is a significant task in news analysis, with substantial research focused on English resources.
• A notable gap exists in resources for analyzing clickbait in non-English news, specifically for Indonesian.
• Existing natural language processing (NLP) research lacks sufficient datasets for Indonesian clickbait classification.

Purpose of the Study:

• To introduce the CLICK-ID dataset, a novel resource for Indonesian news headlines.
• To facilitate research and development in clickbait detection for the Indonesian language.
• To provide a benchmark for evaluating Indonesian clickbait classification models.

Main Methods:

• Compilation of 15,000 Indonesian news headlines from 12 distinct online publishers.
• Annotation of headlines with binary labels: clickbait and non-clickbait.
• Development of an Indonesian clickbait classification model utilizing the created dataset.

Main Results:

• The CLICK-ID dataset comprises 15,000 annotated Indonesian news headlines.
• A clickbait classification model was developed, demonstrating favorable performance on the CLICK-ID dataset.
• The dataset supports replicable experiments in clickbait detection and other NLP tasks.

Conclusions:

• The CLICK-ID dataset significantly contributes to the availability of resources for Indonesian NLP.
• This resource is expected to advance research in clickbait detection and classification for Indonesian news.
• The dataset serves as a valuable tool for future NLP research in under-resourced languages.