Modelling Lexical Ambiguity with Density Matrices

Abstract

Words can have multiple senses. Compositional distributional models of meaning have been argued to deal well with finer shades of meaning variation known as polysemy, but are not so well equipped to handle word senses that are etymologically unrelated, or homonymy. Moving from vectors to density matrices allows us to encode a probability distribution over different senses of a word, and can also be accommodated within a compositional distributional model of meaning. In this paper we present three new neural models for learning density matrices from a corpus, and test their ability to discriminate between word senses on a range of compositional datasets. When paired with a particular composition method, our best model outperforms existing vector-based compositional models as well as strong sentence encoders.

Foundations for Near-Term Quantum Natural Language Processing

“…Similar questions had been asked earlier in quantum foundations when considering causal inference within the quantum context, mostly be Leifer and Spekkens [74,75,43,3]. The recent paper [83] presents three neural models for learning density matrices from a corpus.…”

“…Soon after the development of the quantum model of language it was put to the test, on classical hardware of course, and the model greatly outperformed all other available models for a number of standard NLP tasks [57,65]. Despite the growing dominance of machine learning, successes have continued until recently [114,83,115]. With the 1st DisCoCat paper appearing in 2008 in a somewhat obscure venue [25], independently Baroni and Zamparelli also proposed the adjective-noun model of Section 2.1, again strongly supported by experimental evidence [10].…”

“…More recently there has been increased activity in using density matrices for DisCoCat-NLP [39,76,77,48,83]. Some of these papers address issues that are of foundational physics interest, for example in [39,76,48] the question is asked how to update a density matrix representing a word, given another density matrix, representing an adjective.…”