KaWAT: A Word Analogy Task Dataset for Indonesian
This work addresses the lack of resources for evaluating word embeddings in Indonesian, but it is incremental as it applies existing methods to a new dataset.
The authors introduced KaWAT, a new word analogy task dataset for Indonesian, and evaluated existing pretrained word embeddings on it and downstream tasks, finding that these embeddings reduced training epochs or yielded significant performance gains.
We introduced KaWAT (Kata Word Analogy Task), a new word analogy task dataset for Indonesian. We evaluated on it several existing pretrained Indonesian word embeddings and embeddings trained on Indonesian online news corpus. We also tested them on two downstream tasks and found that pretrained word embeddings helped either by reducing the training epochs or yielding significant performance gains.