Thursday, December 4th at 14:00
Abstract: Syntax and word meaning contribute to the compositional construction of the meaning of sentences. In studying compositional distributional semantics, we discovered that generally models based on vectors tend to forget the syntactic structure of the sentences. Then, we started to investigate how we can partially preserve syntactic structures in vectors. We came across an interesting idea. We discovered a way to represent tree structured data in small vectors.
In this talk, I report on this idea that we called distributed tree kernels (DTK). DTKs are a novel method to reduce time and space complexity of tree kernels. Using a linear complexity algorithm to compute vectors for trees, we embed feature spaces of tree fragments in low-dimensional spaces where the kernel computation is directly done with dot product. We show that DTKs are faster, correlate with tree kernels, and obtain a statistically similar performance in two natural language processing tasks. I finally describe how to use distributional semantics in distributed tree kernels.