Quantum Methods for Managing Ambiguity in Natural Language Processing

TL;DR

This work extends Quantum Natural Language Processing by modeling linguistic ambiguity as probability distributions over syntactic processes rather than solely over lexical meanings. Grounded in the DisCoCat and DisCoCirc frameworks, meanings are represented as density matrices, enabling mixtures and updating via density-matrix operations; the authors map these probabilistic wirings to quantum circuits using an IQP-ansatz, and validate the concept with a preliminary NISQ-style demonstration. Key contributions include formalizing probability distributions over different syntactic connections, extending to dynamic updates in DisCoCirc, and showing via a small dataset that entropy and fidelity metrics can quantify ambiguity and its resolution in quantum-circuit representations. The approach offers a white-box, interpretable alternative to black-box models and provides a pathway toward scalable quantum-semantic reasoning about complex linguistic phenomena like pronoun resolution and verb-phrase ellipsis.

Abstract

The Categorical Compositional Distributional (DisCoCat) framework models meaning in natural language using the mathematical framework of quantum theory, expressed as formal diagrams. DisCoCat diagrams can be associated with tensor networks and quantum circuits. DisCoCat diagrams have been connected to density matrices in various contexts in Quantum Natural Language Processing (QNLP). Previous use of density matrices in QNLP entails modelling ambiguous words as probability distributions over more basic words (the word \texttt{queen}, e.g., might mean the reigning queen or the chess piece). In this article, we investigate using probability distributions over processes to account for syntactic ambiguity in sentences. The meanings of these sentences are represented by density matrices. We show how to create probability distributions on quantum circuits that represent the meanings of sentences and explain how this approach generalises tasks from the literature. We conduct an experiment to validate the proposed theory.

Figures (25)

• Figure 1: A DisCoCat diagram, composing tensors representing meanings of words, guided by grammar
• Figure 2: The meaning of the sentences Mary eats. Mary is hungry. as a pregroup diagram, after pronounResolution who use a different framework based on projections from Fock space.
• Figure 3: The Bloch sphere visualisation of qubit state vectors, the black dot represents the state vector $\ket{\psi}$ of a qubit on the Bloch sphere
• Figure 4: Example circuit encoding the meaning of the sentence Alice plays guitar. Qubits are represented by vertical lines rather than horizontal lines in the usual quantum circuit notation, to emphasise the connection to DisCoCat diagrams. The combination of Hadamard-, CNOT-, and measurement gates correspond to the cup-shaped wires in DisCoCat diagrams.
• Figure 5: The diagram encoding the meaning of the sentence Alice plays …, where the three dots indicate that this word is not available