Near-Term Advances in Quantum Natural Language Processing

TL;DR

The paper demonstrates that early NLP tasks can be implemented on quantum hardware using small datasets, exploring word-based and embedding-based topic classification, bigram modelling via quantum probability, and ambiguity resolution in composition. It compares explicit qubit-per-word/ topic schemes with densely encoded word embeddings, evaluates on lambeq and IMDb-like data, and investigates a Quantum Circuit Born Machine for joint distributions alongside experimental and simulated results. Hardware realities and scaling challenges are analyzed, showing statistically meaningful results on real data but highlighting difficulties in predicting performance from artificial language benchmarks. The work argues for a spectrum of quantum NLP approaches, discusses practical tradeoffs, and emphasizes hardware progress as the key enabler for larger-scale, real-world applications in the near term.

Abstract

This paper describes experiments showing that some tasks in natural language processing (NLP) can already be performed using quantum computers, though so far only with small datasets. We demonstrate various approaches to topic classification. The first uses an explicit word-based approach, in which word-topic scoring weights are implemented as fractional rotations of individual qubit, and a new phrase is classified based on the accumulation of these weights in a scoring qubit using entangling controlled-NOT gates. This is compared with more scalable quantum encodings of word embedding vectors, which are used in the computation of kernel values in a quantum support vector machine: this approach achieved an average of 62% accuracy on classification tasks involving over 10000 words, which is the largest such quantum computing experiment to date. We describe a quantum probability approach to bigram modeling that can be applied to sequences of words and formal concepts, investigating a generative approximation to these distributions using a quantum circuit Born machine, and an approach to ambiguity resolution in verb-noun composition using single-qubit rotations for simple nouns and 2-qubit controlled-NOT gates for simple verbs. The smaller systems described have been run successfully on physical quantum computers, and the larger ones have been simulated. We show that statistically meaningful results can be obtained using real datasets, but this is much more difficult to predict than with easier artificial language examples used previously in developing quantum NLP systems. Other approaches to quantum NLP are compared, partly with respect to contemporary issues including informal language, fluency, and truthfulness.

Figures (9)

• Figure 1: Comparing Quantum (left) with Boolean Logic (right). In quantum logic concepts are modelled by subspaces such as lines and planes, and a point can be projected onto any subspace. In classical logic concepts can be represented by any set of points, and a point either belongs to this set or it does not.
• Figure 2: Basic quantum logic gate diagrams used throughout these examples. A single qubit rotation gate (left) manipulates superposition of $\lvert0\rangle$ and $\lvert1\rangle$ states for the qubit. The two-qubit CNOT gate (right) entangles two qubits
• Figure 3: Example Adder Circuit that Combines the Angles $\theta$ and $\varphi$
• Figure 4: Example Classification Circuit for Two Words and Two Topics
• Figure 5: SVM Kernel Circuit Example with 8-Dimensional Inputs $\vec{x}$ and $\vec{z}$