Experimental Support for a Categorical Compositional Distributional Model of Meaning

TL;DR

This paper implements and evaluates a categorical compositional distributional model of meaning, demonstrating its effectiveness in sentence understanding tasks and showing improved performance with syntactic complexity.

Contribution

It provides the first empirical implementation of the categorical compositional model using real data and evaluates its performance on sentence disambiguation tasks.

Findings

Model matches existing methods on intransitive sentences

Model outperforms competitors on transitive sentences

Performance improves with increased syntactic complexity

Abstract

Modelling compositional meaning for sentences using empirical distributional methods has been a challenge for computational linguists. We implement the abstract categorical model of Coecke et al. (arXiv:1003.4394v1 [cs.CL]) using data from the BNC and evaluate it. The implementation is based on unsupervised learning of matrices for relational words and applying them to the vectors of their arguments. The evaluation is based on the word disambiguation task developed by Mitchell and Lapata (2008) for intransitive sentences, and on a similar new experiment designed for transitive sentences. Our model matches the results of its competitors in the first experiment, and betters them in the second. The general improvement in results with increase in syntactic complexity showcases the compositional power of our model.