Kernelized Sorting for Natural Language Processing

We further develop Object matching, or alignment, is an underlying problem a semi-supervised "bootstrapping" variant of kernelized for many natural language processing tasks, including document sorting that addresses the problem of noise. We compare alignment (Vu, Aw, and Zhang 2009), sentence alignment kernelized sorting with matching canonical correlation (Gale and Church 1991; Rapp 1999) and transliteration analysis (MCCA) (Haghighi et al. 2008) on a wide variety mining (Hermjakob, Knight, and Daumé III 2008; of tasks and data sets and show that these strategies are Udupa et al. 2009). For example, in document alignment, sufficient to turn kernelized sorting from an approach with we have English documents (objects) and French documents highly unpredictable performance into a viable approach for (objects) and our goal is to discover a matching between NLP problems.