First Christmas as a Doctor [5.1.2010]

Hanna Suominen, researcher at the University of Turku, received her best Christmas present, a doctoral degree, ahead of time. On 15 December, Ms Suominen defended her doctoral thesis in which she presents five machine learning applications she has developed for improving the flow of information in health care.

A fresh Ph.D., researcher of information sciences Hanna Suominen is apparently used to the fact that there are always people unfamiliar with information technology among us. She explains with accustomed patience to the layman what machine learning actually means.

− In ordinary computer programming all commands are entered explicitly. In machine learning, instructions have been encoded in the program and by reading them the program will learn to proceed in a specific way, for example to direct certain e-mail messages to deleted items, Ms Suominen summarises.

So the computer learns gradually what is junk mail and what is not. It understands. According to Ms Suominen, machine learning can be compared to the way in which a child learns one thing at a time to tell what is right and what is wrong.

Language search for those who need it

Ms Suominen developed in her doctoral study five machine learning applications by which she aims at facilitating the flow of health information.

She reminds that clear and comprehensive recording of treatment is a legal obligation. The result is a big pile of health records containing largely unorganised text which is not always understandable even to the professionals who need it.

Ms Suominen wants to make it easier to handle that information.

− The aim was to integrate the human language technology searches into health care services. When the staff needs information about a patient's lungs, they could do a search, for example, on ‘breathing', and the language search would give an overall view about the matter from the patient records.

The applications developed by Ms Suominen make the archive search already used in many other fields available to health care staff. The biggest difference is that the language used in patient records is very difficult specialised language of which Ms Suominen herself did not understand much at first.

Problems are caused, for example, by the compact nature of the text. In health records the subject matters change very frequently, on average at eight word intervals, and the language is very difficult to interpret. For example the word ‘noradrenaliini' (noradrenaline) has hundreds of various spellings including misspellings, case endings and abbreviations.

In machine learning these special characteristics of language are understood and all spellings can be included in the search results.

− There are a couple of ways to see beyond spelling mistakes. For example by looking at the neighbouring words, Ms Suominen explains.

People don't know how to utilise machine learning

Finland is often referred to as a country of high technology which has not, however, managed to integrate the state-of-the-art technology into services. This is exactly why Finland's position in different rankings of the development of information society has gradually fallen.

Ms Suominen confirms that in the area of e.g. health records the opportunities of machine learning have not been utilised at all.

− The providers of health care services don't even have simple machine learning programs, not even proof-reading programs. It is an interesting question why these highly developed and advanced programs have not taken their place in everyday routines, she says.

According to Ms Suominen's future vision her applications will be standardised to support the flow of information in the Finnish health care sector. She says that it would save the time of health care staff and improve the quality of treatment, as the health records would become easier to understand.

− Instead of leafing through text, health care professionals could spend their time on their field of special expertise, Ms Suominen summarises.

Model youth of the state

Ms Suominen is young for a Doctor of Philosophy, only 28 years. Many of her age mates have not finished their Master's theses. However, Ms Suominen denies being a child prodigy.

− By concentrating and creating a balance between work and daily life you'll find the time, Ms Suominen says modestly about her achievement.´

She points out that she did not have to perform miracles to get a doctoral degree at the age of 28; it was enough to make it within the target schedule.

The Master's degree took five years to finish and the doctoral thesis four years.

− This is what the government is aiming at. They want young people to graduate within the target schedule to get them quickly from training to working life, Ms Suominen says.


Hanna Suominen

- Born in Loimaa in 1981 - Matriculation examination at Loimaan lukio in 2000 - Master of Science at the University of Turku in 2005 - Doctor of Philosophy at the University of Turku in 2009 - Researcher in the University of Turku and coordinator of the IKITIK (Information and language technology for health information and communication) consortium - Hobbies: needlework, cooking, as well as ballet, jazz dance and salsa