lambeq: An Efficient High-Level Python Library for Quantum NLP

Kartsaklis, Dimitri; Fan, Ian; Yeung, Richie; Pearson, A. W.; Lorenz, Robin G.; Toumi, Alexis; Felice, Giovanni de; Meichanetzidis, Konstantinos; Clark, Stephen R. L.; Coecke, Bob

doi:10.48550/arxiv.2110.04236

Cited by 20 publications

(33 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For example, we could have the meanings of a circuit as a given, and from these deduce the meanings of the words in it. This is in fact what we did in our QNLP-work [90,28,87,76], obtaining the word-meanings from a corpus of sentences we knew to be truthful. The reason for doing so in that particular case is that currently there is no better way available to load classical data on a quantum computer.…”

Section: Top-down Flow and Ambiguitymentioning

confidence: 69%

“…This is an important feature about of shared compositional structure: it enables one to map seemingly entirely different areas onto each other, and enables an exchange of concepts and tools. In fact, in this particular case it gets even better, as this connection effectively enabled us to process natural language on a quantum computer [117,89,90,28,87,76]. For doing so, a language diagram, now interpreted as a quantum process, still needs to be compiled so that it fits on quantum hardware.…”

Section: Examples and Non-examples 51 Natural Language Processingmentioning

confidence: 99%

“…The crux was the fact that quantum compositional structure is, category-theoretically speaking, exactly the same as grammatical structure [21]. This correspondence has recently resulted in implementations of NLP on existing quantum computers, by ourselves at Cambridge Quantum [28,29,90,87,76]. We provide more details on this in Sections 5.1 and 6.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Compositionality as we see it, everywhere around us

Coecke¹

2021

Preprint

Self Cite

View full text Add to dashboard Cite

There are different meanings of the term "compositionality" within science: what one researcher would call compositional, is not at all compositional for another researcher. The most established conception is usually attributed to Frege, and is characterised by a bottomup flow of meanings: the meaning of the whole can be derived from the meanings of the parts, and how these parts are structured together.Inspired by work on compositionality in quantum theory, and categorical quantum mechanics in particular, we propose the notions of Schrödinger, Whitehead, and complete compositionality. Accounting for recent important developments in quantum technology and artificial intelligence, these do not have the bottom-up meaning flow as part of their definitions.Schrödinger compositionality accommodates quantum theory, and also meaning-as-context. Complete compositionality further strengthens Schrödinger compositionality in order to single out theories like ZX-calculus, that are complete with regard to the intended model. All together, our new notions aim to capture the fact that compositionality is at its best when it is 'real', 'non-trivial', and even more when it also is 'complete'.At this point we only put forward the intuitive and/or restricted formal definitions, and leave a fully comprehensive definition to future collaborative work.

show abstract

Section: Top-down Flow and Ambiguitymentioning

confidence: 69%

Section: Examples and Non-examples 51 Natural Language Processingmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Compositionality as we see it, everywhere around us

Coecke¹

2021

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…When considering the software tools available today for designing quantum algorithms, there are many different options but the two classes of tool kits discussed here are generalpurpose established libraries like IBM's qiskit 1 and the niche software stack designed specifically for QNLP, namely lambeq [22]. lambeq is a Python library (released in October 2021) built for QLNP experiments and is the first of its kind.…”

Section: Methodsmentioning

confidence: 99%

Design and Implementation of a Quantum Kernel for Natural Language Processing

Wright¹

2022

Preprint

View full text Add to dashboard Cite

Natural language processing (NLP) is the field of computer science that attempts to make human language accessible to computers, and it relies on developing a mathematical model to express the meaning of symbolic language. One such model, referred to as DisCoCat, defines how to express both the meaning of individual words as well as the compositional nature of the sentence specified by grammar. This category-theoretic model is costly to implement on a classical computer but has an equivalent quantum mechanical formulation, which has lead to its implementation on quantum computers and the creation of the field quantum NLP (QNLP). This recent experimental work used quantum machine learning techniques to learn a mapping from text to class label using the expectation value of the quantum encoded sentence. Some theoretical work has been done on computing the similarity of sentences using this quantum model but they rely on an unrealized quantum memory store. The main goal of this thesis is to leverage the DisCoCat model to design a quantum-based kernel function that can be used by a support vector machine (SVM) for NLP tasks. The kernel function must compute the similarity of two given sentences with high accuracy, efficiency, and resilience to noise. Two similarity measures were studied: (i) the transition amplitude approach and (ii) the SWAP test. A simple NLP meaning classification task from previous work was used to train the word embeddings and evaluate the performance of both models. The Python module lambeq and its related software stack was used for implementation. The explicit model from previous work was used to train word embeddings and achieved a testing accuracy of 93.09 ± 0.01%. It was shown that both the SVM variants achieved a higher testing accuracy of 95.72 ± 0.01% for approach (i) and 97.14 ± 0.01% for (ii). The SWAP test achieved a slightly higher accuracy and yielded a more appropriate similarity matrix. The SWAP test was then simulated under a noise model defined by the real quantum device, ibmq_guadalupe. The explicit model achieved an accuracy of 91.94 ± 0.01% while the SWAP test SVM achieved 96.7% on the testing dataset, suggesting that the kernelized classifiers are resilient to noise. These are encouraging results and motivate further investigations of our proposed kernelized QNLP paradigm. AcknowledgementsI would like to thank my supervisor, Dr. Stephen Hughes, for his insight and guidance throughout this project. This field was new to both of us so I appreciate him for wondering into the unknown with me.As this is the culmination of my undergraduate degree, I would also like to thank my family, friends, peers, and the Queen's Faculty that have all supported me and played a part in the development of my personal and professional self. i Contents List of Tables

show abstract

“…In the circuit model, individual gates, typically defined in terms of their matrices, are presented as arbitrary building blocks that the user is expected to compose into meaningful (and -please -fast) algorithms. Unfortunately, most of us have come to the realisation that this is a highly challenging task, for all but a few applications that themselves present clear "quantum-like" structure [13][14][15][16].…”

Section: M Ementioning

confidence: 99%

Quantum Circuits in Additive Hilbert Space

Luca¹

2021

Preprint

View full text Add to dashboard Cite

Representations of quantum computations are almost always based on a tensor product ⊗-structure. This coincides with what we are able to execute in our experiments, as well as what we observe in Nature, but it makes certain familiar quantum primitives convoluted. Reversible classical circuits, diagonal operations or controlled unitaries all have very elegant and simple matrix representations that cannot be expressed succinctly as a circuit in a simple gate set, complicating quantum algorithm design and circuit optimization.We propose a new additive presentation of quantum computation to address this. We show how conventional circuits can be expressed in the additive space and how they can be recovered. In particular in our formalism we are able to synthesize high-level multi-controlled primitives from low-level circuit decompositions, making it an invaluable tool for circuit optimization. Our formulation also accepts a circuit-like diagrammatic representation and proposes a novel and simple interpretation of quantum computation. C 1. Circuits and Matrices are not Enough . 1 2. Background.

show abstract

lambeq: An Efficient High-Level Python Library for Quantum NLP

Cited by 20 publications

References 25 publications

Compositionality as we see it, everywhere around us

Compositionality as we see it, everywhere around us

Design and Implementation of a Quantum Kernel for Natural Language Processing

Quantum Circuits in Additive Hilbert Space

Contact Info

Product

Resources

About