Teaching a computer human languages

05.04.2011
Tall, blonde and with calm hazel eyes, Alyona Medelyan doesn't strike you as your typical software engineer. Ukrainian-born Medelyan has led the developer team at search software company Pingar for the past 18 months. Now, she has been appointed chief research officer at the company.Pingar develops semantic search solutions to help companies find useful search results from masses of unstructured data. Medelyan's niche expertise in Natural Language Processing (NLP) will drive the start-up's next steps into the future. NLP is about designing programmes that "understand" human language, says Medelyan, who herself speaks Russian, Ukrainian, English, German, some Italian and is learning Chinese."I find it very exciting," she says. "Languages are fascinating and complex -- developing computer applications that attempt to understand them is very challenging. There are still many unsolved problems in this field. New algorithms and new powerful methods are developed every day."So far, she has developed an enhanced summarisation tool to boost the search software. Pingar's software basically pulls out sentences from a document that contain a particular search term and collates them into a PDF-file. But some of the extracted sentences may be out of context, so Medelyan came up with a new algorithm that summarises the document. She has also been in charge of the recent release of an API (application programming interface), which gives third-party developers access to Pingar's core technologies. "I'm very excited to see what other developers will do with it and how the new solutions will compare against other tools in the market," she says, at an interview at Pingar's newly-opened Auckland city office.Not so long ago, the University of Waikato PhD graduate was looking out over another famous skyline. Medelyan's supervisor at Waikato, Ian Witten, had also been the supervisor for another bright mind, Craig Nevill-Manning, who is now director of Google Engineering in New York. Medelyan got in contact with the intern supervisor there who suggested she come over and implement her PhD work, using Google technology. The three-month internship at Google's Manhattan office was a "bit of a change" from Hamilton, she says."I had to buy new clothes to fit in," she laughs. "It was definitely a busy life over there. Almost every night I attended cool events -- concerts, roof-top parties and even yoga classes in the studio of my favourite painter."The working environment was amazing, she says. "There are so many smart people there. Everyone you talked to had won national competitions in mathematics or graduated from MIT with top marks. It was a very inspiring environment."During her PhD (which she finished in 2009), Medelyan developed an open-source tool, Maui (multi-purpose automatic topic indexing), which automatically identifies the main topics in documents. Her tool is now used by the Food and Agriculture Organisation of the UN, a US bank and a search engine in Australia, among others. Being a young woman in a very male-dominated industry hasn't bothered her at all, she says. "There aren't many female role-models [in the ICT industry] so it can be hard to associate yourself with somebody who has achieved a lot in this area; it's much easier for guys to do that," she says."Hopefully I can help break the pattern and help coach other women in the industry," she says.