Artificial intelligence needed to make sense of the heaps of big data information
Big data is widely available these days. To help turn this into solutions, further progress in artificial intelligence is needed, says Evangelos Kanoulas, professor of Text Analytics & Information Retrieval. He intends to do his part, with a focus on applications in the legal domain.
The amount of data being collected in the world continues to rise. To Evangelos Kanoulas, who was appointed professor of Text Analytics & Information Retrieval at the UvA’s Faculty of Economics and Business (FEB) in February, there is a downside to the current focus on gathering ever more data. ‘Big data itself is more of a problem than a solution these days,‘ he argues. ‘One needs artificial intelligence to bring about a solution: computer programmes that consume data and produce valuable information.’
Making sense out of data overload
Building systems that make sense of masses of data has been at the core of Kanoulas’ work for years. His current focus is on unstructured data - data such as text, images or video, whose characteristics cannot simply be put in spreadsheets and analysed. At the UvA’s Faculty of Economics and Business (FEB), he is building algorithms that combine machine and human intelligence. In these algorithms, information provided by humans augments the machine intelligence and vice versa. ‘Take electronic discovery, for example. The human understands the minute details that make something worth discovering but is unable to locate it amidst all the data or even to know whether it exists. The machine can quickly analyse this data but cannot fully capture the importance to the human. A continuous collaboration is needed’, Kanoulas explains.
Helping people make sense of data is a common theme in the career of Greece-born Kanoulas, who took his first computer class when he was in primary school and obtained a PhD in information retrieval at Northeastern University in Boston in 2009. Since then, he has worked at two universities in the UK and at Microsoft Research as well as at Google Research before joining the UvA’s Informatics Institute in 2014.
Major work to be done and progress ahead
In addition to his work in academic and corporate circles, Kanoulas has also been involved since 2009 in research for the National Institute of Standards and Technology (NIST), an agency of the U.S. Department of Commerce. In cooperation with NIST, he has developed benchmarks for quantifying whether changes to an algorithm would make it more effective. Based on the result of testing new algorithms or proposed changes to existing algorithms, researchers can decide whether or not to introduce or change these algorithms. Kanoulas proudly points out that the benchmarks have already been used by hundreds of research groups worldwide: 'This affects the real economy, the real world.’
Kanoulas is convinced that artificial intelligence is going to have a big impact. ‘But we still need to solve many technical problems, making machines smarter’, he adds.
Problems range from ensuring academic integrity and fighting biases in data and in algorithms, to providing ways for users to understand why an algorithm makes certain decisions. ‘I think that will be achieved by the research community within the next ten years.’ Kanoulas expects to contribute by improving the understanding of human language for the next five to ten years.
Benefit beyond academic circles
While working on his PhD, Kanoulas collaborated with the radiation oncology department of Boston’s Massachusetts General Hospital, and developed a keen interest in the healthcare sector. Cancer cells can move relatively quickly, even during a treatment session. This makes directing radiation beams perfectly to kill cancer cells, but not the healthy cells surrounding them, very challenging, Kanoulas explains. He created an algorithm to predict where the cancer cells would go, helping doctors to hit exactly the right spot. ‘Very, very interesting. And you could see the immediate impact.’
Artificial intelligence systems that can help provide more clarity in healthcare will not only improve individual treatments, but also facilitate major scientific discoveries in the future, he expects.
At the FEB, Kanoulas will focus on another topic he is particularly interested in: applying human intelligence in the legal domain. The professor is married to a lawyer and knows how much of a struggle it can be to pull together the information needed to build a case. ‘The electronic discovery process can be quite laborious’, he says. In this process, electronic data that may include millions of emails are located, read, analysed and so on. ‘Maybe five or ten things out of hundreds of thousands items may be of interest.’ Algorithms that provide an effective filter can make a huge difference.
Aided by two PhD students starting in September, Kanoulas will be working on systems to make the due diligence process involved with mergers and acquisitions more efficient. The use of improved algorithms in these ‘virtual data rooms’ should enable rapid screening, analysis, comparison and matching of details in the documents of the companies involved. One of the PhD students will be funded in part by virtual data room services provider Imprima.
Best of both worlds
Kanoulas, who enjoyed doing research at Google and Microsoft, sees clear advantages to working with or for companies. Some things are simply better in the commercial sector, he says: ‘Financial resources, data resources, computational resources… Because of the high salaries, they can attract smart people, and they have real life problems to work on.’ A drawback, however, can be the ‘strong connection to the products.’
For now, he has embraced the academic world. ‘In academics I have more opportunities to learn things. And it is easier to try out a new direction in your research if you see something interesting or exciting happening.‘ Education is also interesting, says Kanoulas, who currently supervises several PhD students and teaches five courses a year. Next to his work at FEB, he keeps the assistant professor post at the Informatics Institute he accepted in 2014.
Meanwhile, he continues to champion the exchange of ideas with business circles. It is one of the reasons for his involvement in Amsterdam Data Science (ADS), which was set up by a number of organisations including the UvA and the VU University Amsterdam to focus on interdisciplinary IT-research in the field of data science. Earlier this year, Kanoulas spoke at a Meetup on Data Science in Legal Tech organised by ADS, along with speakers from Dutch law firms such as De Brauw Blackstone Westbroek. He is keen to combining best of both worlds: 'We're looking for collaborations.’