INARTE: An Indonesian Dataset for Recognition Textual Entailment

Recognition Textual Entailment (RTE) try to solve variability problem that commonly encountered in natural language-based systems. The basic idea is to detect whether the meaning of a text can be inferred by another text. The need dataset in language other than English is necessary to accelerate research development in RTE. We created RTE dataset for Indonesian by retrieval text from Web and generate text-hypothesis pairs as many as possible. The subset technique is used to decide whether Text (T) entails Hypothesis (H). The initial data used 400 question-answer pairs obtained 1,577 entailment pairs, where 481 entailment pairs obtained from the accuracy above 50%.