You wouldn’t use an academic search engine to look for cat videos — but if there’s a video with cats in it that goes with an academic paper, the latest version of Semantic Scholar just might find it for you.
Semantic Scholar is the AI-based search engine that’s been developed by the Seattle-based Allen Institute for Artificial Intelligence, or AI2, specifically to sift through research for the most relevant results.
Over the past three years, the project has indexed more than 40 million research papers. Now AI2’s researchers have turned their algorithms loose to link those papers to associated presentation slides, Github code libraries, summaries of clinical trials, news articles, blogs, social-media postings and videos.
That includes a video with pictures of cats and dogs in it, tied to a paper titled “Unpaired Image-to-Image Translation Using Cycle-Consistent Adversarial Networks.”
It needs to be said that the goal of Semantic Scholar’s latest enhancement isn’t to find geekier cat videos.
“The core problem is what I would describe as information overload,” Douglas Raymond, general manager for Semantic Scholar at AI2, told GeekWire.
Raymond pointed out that since World War II, the volume of scientific literature has been doubling every nine years. That volume is currently growing at the rate of 2.5 million papers per year.
To cut through the overload, information specialists have designed a wide range of algorithms to identify the most influential and impactful papers — based on how much they’re cited by other papers, as well as more subtle factors such as their centrality to the research that follows. “The difficulty with that approach is that it doesn’t work for new science,” Raymond said.
The bonus resources served up by Semantic Scholar — including that cat-and-dog video — can provide additional context, especially for research that hasn’t yet been fully digested by the scientific community.
If a particular paper happens to spark a lot of news articles and blog posts, that doesn’t affect Semantic Scholar’s ranking of the paper’s scientific quality. But those extra resources could help researchers using Semantic Scholar get a better understanding of the paper’s scientific point.
For example, a recently published study traced the impact that sugar and artificial sweeteners could have on strokes or dementia. The study hasn’t been out long enough to spark follow-up research, but Semantic Scholar serves up dozens of reports specifically about the findings.
“We are crossing the chasm between academic papers and more popular media to facilitate a new and smarter way to do science,” Oren Etzioni, AI2’s CEO, explained in a news release.
The project serves as one of the legacies of Microsoft co-founder and philanthropist Paul Allen, who passed away last month. Semantic Scholar is free to use, in keeping with Allen’s intent to use artificial intelligence for the common good.
“By using artificial intelligence to sift through this information overload, we’re removing a key barrier to making science more impactful,” Raymond said. “And we’re trying to increase the impact of every individual scientist.”
More from GeekWire: