Quantum Natural Language Processing
A sentence is not just a “bag of words”,¹ but rather, a kind of network in which words interact in a particular fashion. ...lead to a graphical representation of how the meanings of the words are combined to build the meaning of a sentence as a whole, as opposed to treating the sentence as a structureless “bag” containing the meanings of individual words.
This flow of words in sentences can, in fact, be traced back to work originally started in the 1950s by Chomsky and Lambek among others, that unified the grammatical structures, essentially of all languages, within a single mathematical structure. In particular, the network of a sentence’s meaning-flow is constructed according to a compositionalmathematical model of meaning (semantics).¹⁶
In summary, a direct correspondence was established on the one hand between the meanings of words and quantum states, and on the other hand grammatical structures and quantum measurements.
The first conference on Quantum Natural Language Processing, or “QNLP”as we’ve called it, took place in Oxford in December 2019,¹² where we presented a simulation of our experiment¹⁴ (all talks can be found online¹³).
...by employing quantum machine learning we do not directly encode the meanings of words, but instead construct a framework in which quantum states and processes learn their meanings directly from text.
Naturally, we next turned our attention toward the design and execution of an experiment that is non-trivial, not least of all since our design is predicated on the program being scalable. This means that the dimension of the meaning space grows significantly with the number of qubits available whilst the size of the circuits dictated by the grammar does not grow too large with the size of the sentence.
What's next? ...
Thirdly, and importantly, rather than being restricted to single sentences as was the case in this demonstration, we will process larger text.¹¹ We will provide further information in subsequent articles and papers as the team completes new experiments.
Fourthly, we could embark on other tasks besides question-answering, such as language generation, summarization, etc.
Finally, of course, when hardware becomes more powerful we can simply scale up the size of the meaning spaces and complexity of the tasks — which is clearly our overall objective.
See the full story here: https://medium.com/cambridge-quantum-computing/quantum-natural-language-processing-748d6f27b31d