Multi-Step Regression Learning for Compositional Distributional\n Semantics

Abstract

We present a model for compositional distributional semantics related to the\nframework of Coecke et al. (2010), and emulating formal semantics by\nrepresenting functions as tensors and arguments as vectors. We introduce a new\nlearning method for tensors, generalising the approach of Baroni and Zamparelli\n(2010). We evaluate it on two benchmark data sets, and find it to outperform\nexisting leading methods. We argue in our analysis that the nature of this\nlearning method also renders it suitable for solving more subtle problems\ncompositional distributional models might face.\n

Ähnliche Arbeiten