What is Quantum computing? It is the collective properties of quantum states, such as superposition and entanglement, to carry out the computation. The devices that carry out quantum computations are known as quantum computers. In recent news, Honeywell is coming up with the actual quantum computer, but in order to carry out anything with a quantum computer, one requires software. This is where Cambridge comes to the rescue as it is one company that has invested a ton of time and conducted a lot of research, to build software. There is a trend of big companies getting their hands dirty when it comes to Quantum Computing. These companies are working on quantum computing and have a mindset where they are to able reduce mistakes, increase accuracy and compute the data to use it practically. The conclusion the experts came to was in order for them to use a Quantum Computer, it must have a symbiotic relationship with a classical computer to create different outcomes.
This is when Cambridge announced some software that is usable in the current scenarios in respect to Quantum Computing that can be applied to various businesses. They showcased the QNLP, or Quantum Natural Language Processing Toolkit and Library. Cambridge decided to call it ‘lambeq’, and is the first software toolkit for quantum natural language processing that can take sentences and transform them into a quantum circuit. Cambridge also announced this toolkit is released open-source so that anyone who requires the toolkit has the accessibility to use it. Care Quality Commission stated that since this toolkit can run on computers, like IBM’s quantum machine, which is not Honeywell. It is not a tie-up in any way but these circumstances have these 2 companies working together.
Quantum Natural Language Processing uses Sentences as networks. A sentence is not just considered a bag of words, but it works more like a network where different words interact in various situations. 10 years ago two colleagues, Mehrnoosh Sadrzadeh and Steve Clark started to create these networks. This led to a graphical representation of how the meanings of the words are connected to build the meaning of a sentence as a whole, compared to treating the sentence as a structureless bag accommodating the meanings of separate words. A created network would look like this:
Timony loves Josie
In the example above, the subject Timony and the object Josie are both sent to the verb ‘loves’ and together they make up the meaning of the sentence. This flow of words in sentences can be tracked originally started in the 1950s by Chomsky and Lambek amid others, which unified the grammatical structures, of all languages, inside a single mathematical structure. Particularly, the network of a sentence’s meaning flow is structured according to the compositional mathematical model of meaning.
The experimental workflow is explained as: Let’s take A as the grammar category, i.e, the mathematical model where grammar diagrams are created. As stated above, a grammar diagram (or network) enciphers the flow of word meanings in a grammatical sentence. If one further dives into this concept, a diagram is nothing more than the grammatical and syntactic parsing of a sentence depending on the specified grammar model.
Obviously, this led to the question, can one make quantum computers handle natural language? Will Zeng and BC first proposed in a paper in 2016. The paper created a new pedestal for Natural Language Processing (NLP) in a quantum computing context. To know more about NLP, click here. However, this kind of concept wasn’t without its challenges. The major issue that they faced was the fact that there weren’t any sufficiently capable quantum computers that could implement the tasks that were given from the NLP. Additionally, an assumption was made that one could encode word meanings on the quantum computer using quantum random access memory (QRAM), which is still, to this data, only theoretically possible. The idea of making quantum computers process natural language is just not only exciting but also a very natural step to take for various reasons. Many companies are brewing up ideas to make this possibility a reality.
One could essentially tweak the kind of hardware that one uses, for example, ion traps or optics instead of superconducting qubits. By implementing the toolkit, this development will be incredibly fast-paced. One could also vary the computational model, for example, one could use Measurement-Based Quantum Computation (MBQC as the ability to transfer quantum states in unit time to accelerate addition ) instead of circuits. Rather than being confined to single sentences, one could process larger text., we could work on other tasks besides question-answering, such as language generation, summarization, etc. Lastly, when hardware becomes more powerful we can simply scale up the size of the meaning spaces and complexity of the tasks — which is clearly the overall objective.
Want to learn about how Natural Language Processing is used?