On December 10, Cambridge Quantum Computing (CQC) announced that it has developed ‘meaning-aware’ Quantum Natural Language Processing (QNLP) on a quantum computer. This proves that QNLP is fully quantum-native and will be expected to provide near-term advantages over even the fastest classical supercomputers.
In layman’s terms, QNLP allows a system to understand grammatical structure and meanings of words in ways classical computers cannot. Classical computers do not have enough processing power to understand the meaning of both of these. Quantum computers, on the other hand, are capable of doing so.
Natural language processing (NLP) is the current frontier of research in contemporary artificial intelligence (AI). It holds the crown of perhaps the most challenging area of the AI field. In classical computers, achieving ‘meaning-aware’ NLP is but a distant dream and goal.
However, with advances in quantum computing hardware and algorithms, QNLP might not be so far off as originally perceived.
Papers posted on arXiv, a repository of scientific e-print, scientists from CQC have provided the conceptual and mathematical foundations for near-term QNLP. The papers are written in syntax that quantum computer scientists can understand.
Professor Bob Coecke of Oxford University and his team have proven that quantum computers are able to achieve ‘meaning-aware’ NLP, and it is on par with simulating quantum systems. Variational quantum circuits, the leading Noisy Intermediate-Scale Quantum (NISQ) paradigm used to encode classical data on quantum hardware, make NISQ very QNLP-friendly. Prof. Coecke and his team were trying to canonically combine linguistic meanings with rich linguistic structures, particularly grammar.
Previously, CQC was able to speed up QNLP tasks and demonstrate potential quantum advantage for NLP. The ways included using exponentially large quantum state spaces to algorithmically speed up search-related or classification tasks. These are among the dominant tasks within NLP. The large quantum state spaces could accommodate complex linguistic structures. Another way to demonstrate the potential quantum advantage and speed up QNLP tasks was to employ density matrices through novel models of meaning.
The experimental paper that comes with the foundational exposition describes how two powerful IBM quantum computers run the first implementation of an NLP task. This is possible due to CQC having access to the IBM Quantum Network’s powerful resources. Word-meanings are encoded in quantum states and sentences are instantiated as parameterised quantum circuits. The scientists primarily wanted to account for grammatical structure, which is not common in mainstream NLP, through hard-wiring it as entangling operations. The result is a NISQ-friendly approach to QNLP, and it shows promise for scalability.
CQC is exploring potential uses for QNLP, including the possibility of speeding up medical diagnoses, combining an X-ray with a radiologist describing with words what the X-ray is showing in order to identify an illness quickly.
About Cambridge Quantum Computing
CQC was founded in 2014 and enjoys the backing of some globally-acclaimed quantum computing companies. A world leader in quantum software and quantum algorithms, it allows clients to get the most out of rapidly evolving and improving quantum computing hardware. CQC is based in the UK but has offices in the USA and Japan as well, in which over 130 professionals work.