The UKP ASPECT Corpus includes 3,595 sentence pairs over 28 controversial topics. The sentences were crawled from a large web crawl and identified as arguments for a given topic using the ArgumenText system. The sampling and matching of the sentence pairs is described in the paper. Then, the argument similarity annotation was done via crowdsourcing. Each crowd worker could choose from four annotation options (the exact guidelines are provided in the Appendix of the paper).
