NTU Irony Corpus

Introduction

The NTU Irony Corpus consists of more than 1,000 microblog messages collected from the Plurk website. All the messages in the corpus are in Traditional Chinese and have been confirmed to be ironic. They are marked with three types of labels: (1) ironic word/phrase , (2) context, and (3) rhetoric element.

Format

Example:

<context sentiment="pos">才剛買的書,竟然掉頁了,</context>這品質<rhetoric>也太</rhetoric><ironic sentiment="neg">好</ ironic>了<rhetoric>吧</rhetoric>

Tag description:

Download

Click here to download the NTU Irony Corpus.

How to Cite the Corpus

Please cite the following paper when referring to the NTU Irony Corpus in academic publications and papers.

Yi-jie Tang and Hsin-Hsi Chen (2014). “Chinese Irony Corpus Construction and Ironic Structure Analysis.” Proceedings of the 25th International Conference on Computational Linguistics (COLING 2014), 23-29 August 2014, Dublin, Ireland.