Making use of latent semantic analysis, we explore the hypothesis that local linguistic context can serve to identify multi-word expressions that have noncompositional meanings. We propose that vector-similarity between distribution vectors associated with an MWE as a whole and those associated with its constitutent parts can serve as a good measure of the degree to which the MWE is compositional. We present experiments that show that low (cosine) similarity does, in fact, correlate with non-compositionality.

No abstract

We give an in-depth account of compositional matrix-space models (CMSMs), a type of generic models for natural language, wherein compositionality is realized via matrix multiplication. We argue for the structural plausibility of this model and show that it is able to cover and combine various common compositional natural language processing approaches. Then, we consider efficient task-specific learning methods for training CMSMs and evaluate their performance in compositionality prediction and sentiment analysis.

No abstract

The existing search engine interaction paradigm of typing in keywords and getting an enormous list of links is not suited for a lot of information seeking tasks. We see synthesis or summarization of information to satisfy users' information needs as an important step on the way to next generation information access systems. The idea is to explore the alternative geometrical models of meaning [14] based on the theory of Quantum Mechanics and Hilbert Spaces as a unifying framework for integrating semantic metadata into retrieval and summarization.