Extracting Keyphrases from News Articles Using Crowdsourcing

Keyphrase extraction is a very important task in text mining. However, keyphrase extraction of news articles cannot be addressed by existing machine-based approaches effectively because of various reasons. This paper employs crowdsourcing for keyphrase extraction of news articles. We first design a proper crowdsourcing mechanism to extract keyphrases from news articles and then adapt three truth inference algorithms (namely IMLK, IMLK-I, and IMLK-ED) for integrating multiple lists of keyphrases provided by workers. The experiments show that crowdsourcing can significantly improve the performance of the machine-based approach (i.e., KeyRank).