Imperial Graduate School Blog

by Clavance Lim, MSc Student in the Department of Computing

Translating words to numbers

As humans, one way in which we are unique is our ability to communicate with complex language (arguably, science students possess this skill too). In contrast, computers ‘think’ not in language, but in binary numbers. Instead of the decimal system we count with, which uses the ten unique digits ‘0’ to ‘9’, computers ‘think’ only in ‘0’s and ‘1’s. This is because their hardware is controlled by tiny switches, which turn electrical current on or off. As it is difficult to control electrical current at such a microscopic level (switches can be as small as only 10x the size of an atom!), the hardware only works with two states, ‘on’ and ‘off’, which correspond to ‘0’ and ‘1’. So everything we do on a computer – from pressing a single key on the keyboard, to watching a movie – has to be converted to a series of instructions in the form of ‘0’s and ‘1’s.

In recent years, there has been some hype surrounding the pursuit of ‘artificial intelligence’, or the creation of computers or machines to perform tasks requiring human intelligence. To achieve this, any task must be represented in the form of numbers, for the computer to process it. Thus, one question the field of natural language processing faces is: how do we translate words to numbers, while allowing words to retain their linguistic meaning?

A key breakthrough has been to design algorithms which convert each word to a vector (which is simply a row of many numbers). A famous example of the success of this approach is when researchers managed to show that the vectors for ‘Man’ deducted from ‘King’ plus ‘Woman’ resulted in the vector for ‘Queen’.1

Part of my education at Imperial has been to examine the application of this to legal documents, a field in which language is particularly important. For example, we can see that even when mapped to a small grid, words which have similar meanings are placed closer to each other.

Meaningfully representing words as numbers unlocks the potential for computers to do much more. For example, using a much older method,3 the first paragraph of this essay was summarised as:

“Instead of the decimal system we count with, which uses the ten unique digits ‘0’ to ‘9’, computers ‘think’ only in ‘0’s and ‘1’s. So everything we do on a computer – from pressing a single key on the keyboard, to watching a movie – has to be converted to a series of instructions in the form of ‘0’s and ‘1’s.”

This already seems to capture the gist of the paragraph. With current research, the aim is to accurately summarise documents not only by picking out the most important sentences, but by rewriting the entire passage using words unseen in the text itself. From distilling complex articles to designing intelligent chatbots, the potential of this research is tremendously exciting.

References:

1. Mikolov et al. (2013), Efficient Estimation of Word Representations in Vector Space, available at: https://arxiv.org/abs/1301.3781
2. This diagram is from my dissertation, available at:
   https://www.imperial.ac.uk/media/imperial-college/faculty-of-engineering/computing/public/1819-pg-projects/An-Evaluation-of-Machine-Learning-Approaches-to-Natural-Language-Processing-for-Legal-Text-Classi%EF%AC%81cation.pdf
3. Mihalcea and Tarau (2004), Bringing Order into Texts, available at: https://web.eecs.umich.edu/~mihalcea/papers/mihalcea.emnlp04.pdf

Read 4Cs Science Communication Writing Competition – People’s Choice Award in full

by Michelle Lin, MRes Student in the Department of Life Sciences

Cryptococcosis: The Silent Killer

The young patient presented to the hospital with a fever, headache, seizures, and both eyes bulging out of their sockets. Suspecting an infection, doctors first treated the boy with a common antibiotic, Penicillin, presumably to knock out whatever bacterial agent they believed was causing his symptoms.¹

With the boy’s condition failing to improve, doctors kept the boy hospitalized as they searched for a diagnosis and administered various antibiotic and antiviral medications.

As his hospital stay dragged on, the boys condition continued to deteriorate until, after 52 days of ineffective treatments in the hospital, the boy succumbed to his illness. Post-mortem, doctors were able to confirm the boy had been suffering from cryptococcosis, an invasive fungal infection that, without proper anti-fungal treatment, is almost uniformly fatal.

He was six years old.

——————————————————————————–

A fungal infection?

Pathogenic fungi (meaning they are disease causing) are the silent killers of the emerging infectious diseases. Rarer than bacterial and viral infections, invasive fungal infections are often overlooked as a major cause of mortality, while still accounting for approximately 1 million deaths a year.²³ The fungal infection that killed the young boy described above, cryptococcosis, is one of these “silent killers.” Caused by two species of fungi commonly found in the environment, Cryptococcus neoformans and Cryptococcus gattii, cryptococcosis is responsible for upwards of 181,000 deaths per year.⁴

How does infection occur?

Acquired through exposure in the environment, infection can occur years after the initial inhalation of airborne Cryptococcus particles. After traveling through the respiratory tract, these spores settle in the lungs and from there can infect virtually any organ in the body, with the most common targets being the brain, heart, eyes, and lungs. While cryptococcosis infections can be seen in patients with healthy immune systems, the majority of cryptococcosis cases occur in “immunocompromised” populations. ⁵

Yikes- so what is there to do?

With low- and middle-income countries disproportionately affected by cryptococcosis, disease prevention is often the most sensible public health strategy available.⁶ Knowing where Cryptococcus is in the environment gives public health officials the ability to set guidelines for where vulnerable individuals should avoid going and could even prove beneficial for patients by factoring high-risk locations into the differential diagnosis pipeline. The faster a diagnosis is reached with cryptococcosis the better, as timely delivery of anti-fungal medications can be the difference between life and death.

However, all of this requires knowing where C. gattii and C. neoformans are in the environment. That’s where my research comes in. Using field work, statistical modelling, and GIS, I map where in the environment Cryptococcus could be found.

By modelling known presence locations and environmental variables, we are able to uncover the environmental factors most important to each species’ geographic spread and can even create predictive maps depicting where Cryptococcus might be lurking.

By highlighting environmental reservoirs of infection, we are able to determine areas that pose a higher risk of disease transmission in the hopes of one day reducing infections and preventing premature deaths from this environmental scourge.

Read 4Cs Science Communication Writing Competition – 1st Place in full

by David Ho, PhD Student in the Department of Physics

A really strong magnet can dissolve Everything

One wrong thing everyone knows about the universe is “conservation of matter”. It seems obvious: if you have a chair, you can move it, or turn it around, and you still have one chair. If these were the only experiments you did, you might proclaim that the number of chairs in the universe always stays the same.

Of course, it doesn’t take much thought to counter this: with a hammer you can easily change the number of chairs in the universe. But if you collect every splinter of leftover wood, you’ll find the same amount before and after the destruction. Is wood conserved? Of course not; just light a match. But if you count all the atoms…

This example could continue for a while. The main theme is that certain things appear to be conserved, but if you put in enough energy you can break apart the original unit and find a smaller one. The natural question, then, is where does this end? To current physicists’ best knowledge, there are two endpoints to the chain of reduction: all matter is made up of baryons and/or leptons.

The baryons that most people are familiar with are protons and neutrons(1). The most familiar leptons are electrons (less familiar are neutrinos, muons and tauons). Every interaction ever observed conserves baryon and lepton number(2): you can change neutrons into protons, or neutrinos into electrons, but we have never seen a proton become an electron.

Baryon and lepton conservation seems, experimentally, to be unavoidable. But theoretical physicists have found a loophole: a process known as the sphaleron can transform baryons into leptons and vice versa. This is a type of quantum tunnelling: the strange rules of quantum mechanics allow a baryon to pop out of existence and a lepton to take its place. Like nuclear fusion in the lab, sphaleron processes are possible but haven’t been achieved yet. This is because they are phenomenally unlikely: the chance of seeing one is about one in 10160, so small lightning-striking-lottery-ticket analogies aren’t worth the effort.

It turns out, however, that a magnetic field can help this process. In fact, a strong enough magnet would make sphalerons so likely that anything in the magnet’s field would have its baryons converted to leptons, and the object would dissolve. My recent research has been to calculate exactly how strong this is, and we’ve confirmed a long-held suspicion that it’s around 1020 Tesla. Unfortunately (or perhaps thankfully), our strongest permanent magnets are only about 10 T. My funding doesn’t extend to death-ray development, but perhaps one day supervillains will thank 21st Century theorists.

Read 4Cs Science Communication Writing Competition – 2nd Place in full