Understanding travel from web queries using domain knowledge from Wikipedia

Abstract

Developing a deeper understanding of the travel domain is helpful for presenting users with consistent and reliable information, and few sources of data are able to achieve that. Further, such information can serve as background knowledge for evaluating machine learning algorithms. In this paper, we present part of our work towards developing such an understanding. We demonstrate a simple extraction technique and how the extracted data can be used to evaluate an unsupervised embedding model built on search queries with travel intent.