Data Science Investment Counter —Funding raised by UK data science companies in 2018.
£ 5.640 Million

This Search Engine Uses AI to Keep Scientists Updated

For scientists, keeping up with the latest research is essential. But scouring through academic journals every day in search of relevant articles can be extremely time-consuming, while search engines like Google Scholar will just not cut the mustard.

Articles on Google’s academic version are shown on the basis of the number of citations they rake in, so more recent studies get often lost in the noise. Even worse, on those engines it can be daunting to find anything unless one knows exactly what to search for — specific keywords, authors, etc.— a mechanism that makes it harder to come across relevant research one has not heard of.

It was with these problems clear in their minds that Cambridge PhD Vivian Chan and machine learning expert Nilu Satharasinghe set out to create Sparrho.

Launched in 2013, Sparrho is a discovery engine that helps scholars and non-experts alike stay on top of the scientific research that most matters to them. Through personalised feeds, users can be constantly updated on topics they are interested in, with the newest studies sourced from over 45,000 journals and conference proceedings every single day.

Screen Shot 2016-09-07 at 13.50.28

Dr Vivian Chan created AI-driven search engine Sparrho to help scientists stay on top of the latest academic research every day

The system is powered by a blend of artificial and human intelligence— with the former doing about 70 percent of the work, and the remaining 30 left to flesh-and-blood researchers.

“We are relying on researchers flagging up the most relevant articles as they are trained to do this,” Chan says. “In research settings the only metric of success is the number of new publications— that means they have to stay on top on anything else and being sure nobody beats them to it.”

Platform users can collect, recommend and discuss articles on thematic lists or “pinboard.” Sparrho’s team noticed that people who “pin” articles most frequently are usually PhDs, researchers and thought leaders.

Serendipity always leads to innovations, but right now only humans can truly do serendipitous discovery. Using those connections, we are training Sparrho’s engine. – Vivian Chan

“That’s where we get the data points we need to train our AI,” Chan explains.

Relying on that data, Sparrho hopes to teach its AI how to make serendipitous, interdisciplinary connections— that is, showing articles whose relevance to a certain field can be only spotted by an expert eye.

“Serendipity always leads to innovations,” Chan says. “But right now only humans can truly do serendipitous discovery. Using those clever connections  we are training the engine. Over time we hope AI could start doing 80 percent of the work.”

Chan emphasises that, while initially conceived from a scholarly standpoint, Sparrho also aims at bringing science to the larger public.

“ The Royal Society has estimated that the result of poor communications of scientific research— specifically healthcare-related one— affects 100 million individuals, who often end up looking for dubious information on internet forums,” she says.

“With Sparrho we can leverage experts to start posting important studies on pinboards, and possibly creating new pieces of content that would be much more digestible and understandable to the wider community.”

Follow Vivian on Twitter


Co-working space and blog dedicated to all things data science.

Subscribe to our newsletter