Descending-Path Convolution Kernel for Syntactic Structures

Chen Lin, Timothy A. Miller, Alvin Kho, Steven Bethard, Dmitriy Dligach, Sameer Pradhan, Guergana Savova

Abstract: Convolution tree kernels are an efficient and effective method for comparing syntactic structures in NLP methods. However, current kernel methods such as subset tree kernel and partial tree kernel understate the similarity of very similar tree structures. Although soft-matching approaches can improve the similarity scores, they are corpusdependent and match relaxations may be task-specific. We propose an alternative approach called descending path kernel which gives intuitive similarity scores on comparable structures. This method is evaluated on two temporal relation extraction tasks and demonstrates its advantage over rich syntactic representations.