Features

Finding the meaning behind the words we use is something humans are so good at that it often seems simple.

But for computers, understanding the emotions embedded in text is a very difficult task.

I spoke to Stephen Pulman of Oxford University's Computing Laboratory about his research which is helping computers to see what we mean:

OxSciBlog: Why do computers find it hard to understand the meaning behind words?
Stephen Pulman: There are several reasons. Words can be ambiguous: for example, 'cap' can mean 'beat', as in 'I can cap that', or 'limit', as in 'the government will cap student numbers'. If I hear 'the Council is to cap parking charges' I know - or at least hope - that 'cap' means 'limit' in this context. Computers have to be told this kind of thing explicitly, because they don't know anything about the world. Linguistically, the sentence could also mean that the Council were trying to beat some other Council's charges.

Another reason is that frequently we make inferences from words. If I want to know when the new Vice-Chancellor started work in Oxford, and I find a press release saying 'Professor Andrew Hamilton will be ceremonially installed as Oxford's 271st Vice-Chancellor today' then - at least, when I know which date 'today' referred to: another potentially difficult problem -  I know when the new Vice-Chancellor (officially) started work.

OSB: How does your approach help them assess the emotional meaning of text?
SP: We have a very large list of words annotated for the emotional meaning they carry, and we also take account of the grammatical context in which these words occur, so that the effects of negation and other constructs that change meaning can be taken account of. A word like 'progress' is generally perceived as positive, but not when it is in a context like 'fail to progress', or 'little progress'.

OSB: What are the advantages of your approach over existing systems?
SP: By taking account of grammatical context, we can determine emotional attitudes towards the entities and relations mentioned in a text, rather than just characterising the text overall as positive or negative. For example, a movie review might be enthusiastic about the production, but critical of the plot, or praise one actor and criticise another.

OSB: How might systems using your approach be useful to firms and government agencies?
SP: Many large companies are concerned about their reputation, and the reception that their products receive in the marketplace. By analysing
news reports, blogs, or postings on social media like Facebook or Twitter, companies can get almost instant feedback about this. Government agencies can follow the attitudes of dissident or terrorist groups by using systems like ours to track mentions of people or places in intercepted emails and texts or on web sites, particularly when combining our technology with automated translation.

More about this work in an article in The Economist.

Stephen Pulman is Professor of Computational Linguistics at Oxford's Computing Laboratory.

A new study published last week in the New England Journal of Medicine has shown that a simple cooling treatment for babies that suffer a lack of oxygen at birth could avoid brain damage in over 100 babies a year.

The TOBY (Total Body Hypothermia for Neonatal Encephalopathy) trial was led jointly by Oxford University and Imperial College London and was funded by the Medical Research Council.

I talked to TOBY Study Co-ordinator, Brenda Strohm RGN of the National Perinatal Epidemiology Unit [NPEU] at Oxford University.

OxSciBlog: What is birth asphyxia and how can it harm babies?
Brenda Strohm: Birth asphyxia occurs when a baby experiences lack of oxygen during labour and delivery, which can result in damage to the brain and other organs. This can occur if the umbilical cord is tight around the baby or the placenta separates from the womb too early, for example. Asphyxia can occur in both preterm and full term births; the TOBY Study was concerned only with full term babies. It is estimated that about 2 in 1000 full term births are affected, which is about 1400 cases a year in the UK.

OSB: What are the problems in detecting or treating birth asphyxia?
BS: In some cases there is warning that the baby is at risk, with signs of distress in the baby during labour or if there is a haemorrhage for example. But there may be no warning at all; only when the baby is born needing resuscitation does the condition become apparent.

OSB: What did the TOBY trial set out to investigate?
BS: The aim of TOBY was to find out whether whole body cooling (or hypothermia) as a treatment for birth asphyxia improves the outcomes for these babies and is a safe treatment.

The treatment is simple to use: the baby is nursed on a mattress which is cooled by circulating fluid within it. Typically the mattress temperature is set at around 28°C to maintain the baby’s temperature at 33–34°C, instead of the normal 37°C. The cooling lasts for 72 hours and then the baby is slowly re-warmed.

Brain injury from lack of oxygen occurs in two phases: immediately, when the brain is starved of oxygen and then later when the blood supply is restored. There is a chemical cascade when the machinery of the cells breaks down when it is possible to intervene to prevent long-term damage. It is during this ‘therapeutic window’ that cooling treatment is used as the intervention.

OSB: What were the findings of the study?
BS: 325 babies were recruited immediately after birth and were randomly allocated cooling treatment with intensive care or standard intensive care. While there was no difference in the deaths that occurred in the two groups, there was a significant improvement in the rate of survival without neurological abnormality in the cooled group. The cooled group showed improved outcomes in a range of other developmental measures at 18 months of age.

The study is important because birth asphyxia is associated with high morbidity in survivors that is a major burden for the child and family and on health and educational resources as the child matures. Up to now there has not been an effective treatment to offer these babies. Any treatment or intervention that can reduce this burden not only improves quality of life but reduces the costs of providing ongoing care.

OSB: What should pregnant women take from this study?
BS: Fortunately the vast majority of pregnant women will never need to know about cooling for their baby. For those whose baby is affected by birth asphyxia, there is now the knowledge that there is a simple treatment that could make a real difference to their baby’s future, although not every baby is bound to benefit from it.

OSB: What would you like to see happen next based on these findings?
BS: Cooling is not currently considered to be a standard treatment, but clinicians are increasingly using it. In the TOBY Study there were 34 UK neonatal units equipped to offer cooling, now there are nearly 50 with others obtaining the equipment and undergoing preparatory training. Many babies are transferred to units that can provide cooling treatment.

The National Institute for Health and Clinical Excellence (NICE) will be reviewing cooling as a treatment for birth asphyxia in full term babies now that TOBY has published its results. If NICE do recommend cooling, then it will become a standard treatment in the UK.

What are you doing this Friday lunchtime?

If you are NASA's LCROSS spacecraft then you will be crashing into the surface of the Moon.

As Chris Lintott of Oxford's Department of Physics tells The Guardian's Science Weekly podcast this is no accident but exactly what it's been designed to do.

The purpose behind LCROSS's suicide mission is to find out more about water ice deposits hidden in the eternally dark depths of craters on the lunar surface. Finding such deposits is important as water would be a source of the hydrogen and oxygen needed to make rocket fuel as well as a boon for thirsty astronauts.

At 12:30 BST the spacecraft's rocket stage, Centaur, will smash into the lunar crater Cabeus. The LCROSS satellite will be following a few minutes behind so that it can fly through the plume of debris caused by the initial impact, analysing the materials ejected from the crater.

Chris tells us that, if you can get a clear view of the Moon, you should be able to see the glitter as the plume catches the sunlight. Hopefully those monitoring LCROSS's instruments will be able to see more, discovering much about the ancient history of the Moon and our solar system.

In the podcast Chris also gives an update on the Galaxy Zoo peas, discusses dark matter, and shares the excitement generated by the recent discovery of an extrasolar planet that orbits its star backwards.

Could this be evidence for a planet 'kidnapped' from a rival star? Or under the influence of some distant super-Jupiter?

At the moment we don't know, but as Chris comments these discoveries are making us realise just how weird other systems are: 'there's a whole zoo of different solar systems out there!' he tells The Guardian's Nell Boase.

Bioengineering is an exciting and diverse field covering everything from carbon nanotubes to microrobotics and from gene therapy to virtual drug testing.

Oxford’s IBME recently hosted Bioengineering '09, a conference bringing together the latest research in this area.

I caught up with conference organiser Yiannis Ventikos of Oxford University’s Department of Engineering Science to ask him about his group’s research into modelling stents – just one example of what can happen when engineering science meets clinical practice:

OxSciBlog: How does modelling help inform surgical interventions?
Yiannis Ventikos: Uniquely, computational modelling, or ‘simulation’, has the capability to predict the outcome of interventions, in a personalised and patient-specific fashion. Appropriate computer models can answer questions like: ‘how will this disease progress?’, ‘should we intervene aggressively, say surgery, or should we follow a more conservative approach, for example pharmaceutical?’ or ‘if we have decided on intervention, which of the available scenarios is the preferred one?’

The unique power of simulation is that it answers these questions in a personalised fashion. Statistics, demographics and epidemiology of the type commonly used in clinical decision-making are used to validate the models, but once the models are validated, they answer these questions, sometimes from first principles, for the specific genetics, family history and morphology/physiology of each individual patient.

So, the answers obtained are not based on some vague average but on the particular characteristics of the person in need of treatment for the specific disease.

OSB: Why is understanding how different stents work so important?
YV: Stents are used to treat many types of vascular pathologies. A relatively new approach used in interventional neuroradiology is to implant open stents at the neck of cerebral aneurysms in order to reduce blood flow into them. This procedure is a minimally invasive one, since the stents can be delivered using a catheter, through the artery, entering the body via a tiny incision, usually at the thigh.

Oxford and the Department of Neuroradiology are amongst the world’s leading centres where such aneurysms in the brain are treated and therefore making sure that this procedure is done optimally is really important. The open stents used in this procedure are effectively cylindrical-shaped wire meshes that allow blood to pass towards side vessels, but reduce flow adequately towards the aneurysm sack so as to promote stagnation and the formation of a stable clot in the aneurysm.

OSB: How can modelling help to evaluate stents?
YV: Open stents for cerebral aneurysms sound like an interventionist’s dream treatment, a minimally invasive approach that actually addresses the problem for a very wide range of aneurysm morphologies. There is a problem however: not all stents successfully occlude (close) all aneurysms.

The problem here is one of fluid dynamics: putting what is effectively a grid in front of the aneurysm opening must lead to a reduction of inflow that is adequate to induce stagnation and clotting. There is no intuitive way to estimate that however. The reduction of inflow depends on the stent weaving, but also on the local haemodynamics: the speed and direction of blood flow, the morphology and orientation of the opening, and above all, on the interplay of all of the above. Second-guessing whether a stent will work or not is no good – if a wrong decision is made, further interventions are then needed that burden the patient.

Simulation answers this question brilliantly: stents from different manufacturers (or new designs) are virtually placed into position, in a similarly virtual anatomically accurate representation of the vasculature, and their performance is evaluated preoperatively.

Not only can we choose the best device for a particular aneurysm, we can also decide on the best orientation and positioning of the implant. It’s all virtual, done on the computer a day or two before the intervention and it is 100 per cent safe and repeatable. It gives us confidence that when the time for the actual intervention comes, the clinician’s skills will be supplemented with robust knowledge of which device will do the best job for a particular patient, for a particular aneurysm actually, and how this device should be used for best results.

OSB: What are the next steps in your research?
YV: The conceptual framework described above (predictive, simulation-based interventional planning) can be applied to a whole host of conditions. Similarly, the techniques and tools we develop for one disease are often applicable, with little modification, to other diseases too.

We are expanding this technique now to account for a different type of implant, a stentgraft, as used for the treatment of acute aortic disease (like for example, aortic dissections). There, closed conduits (made of reinforced synthetic material) are used to line and so strengthen weak, diseased aortas. Again, models that are as representative of the real situation as possible are being devised; models that incorporate the blood flow dynamics, the vascular wall adaptation and its interaction with the implant, biochemical processes like thrombosis etc.

Over 20 Oxford volunteers recently pitched a tent at this year's Royal Berkshire Show, Newbury, to tempt visitors into exploring scientific experiments and ideas.

The entertainment on offer included mathematical games - such as building a giant fractal - demonstrated by Marcus's Marvellous Mathemagicians, slime-making and burning magic paper performed by chemist Roger Nixon, and a range of smaller experiments out of which the soap and bath bomb-making proved particularly popular.

Access Coordinator and organiser Naomi Capell tells us: 'The main part of the tent was a variety of little science busking experiments, things like creating homemade lava lamps, making coloured non-Newtonian fluid with cornflour, water and food colouring, and exploding cola bottles with mints.'

'It was a very busy day but the feedback we got from parents and teachers was great!'

If you fancy a taste of what was on offer at Newbury then you can join Marcus's Marvellous Mathemagicians tomorrow [1 October] where they'll be running a show called Playing with Numbers at London's Barbican.

Whilst you're at the Barbican why not also go to see their mathematical leader, Marcus du Sautoy, give a talk tomorrow evening on Symmetry: A Journey into the Patterns of Nature [tickets for these events sold separately].