Towards a Matrix-based Distributional Model of Meaning
Eugenie Giesbrecht · 2010
Vector-based distributional models of semantics have proven useful and adequate in a variety of natural language processing tasks. However, most of them lack at least one key requirement in order to serve as an adequate representation of natural language, namely sensitivity to structural information such as word order. We propose a novel approach that offers a potential of integrating order-dependent word contexts in a completely unsupervised manner by assigning to words characteristic distributional matrices. The proposed model is applied to the task of free associations. In the end, the first results as well as directions for future work are discussed. 1