Before embarking on a new research project, a thorough and exhaustive review of existing literature must be done to make sure the new project is novel. Researchers can also explore the entire body of previously published data on a subject to answer a new question using that same data. This is a daunting task, especially considering that millions of new research articles are published each year. Where does one even begin to explore all of that data? This new software may help.
It is hard to get people excited about research software says Eliza Grames, a PhD candidate in ecology and evolutionary biology. Yet, the software she has developed is exciting and to understand why, it is important to put yourself into the shoes of a researcher.
Before embarking on a new research project, a thorough and exhaustive review of existing literature must be done to make sure the new project is novel. Researchers can also explore the entire body of previously published data on a subject to answer a new question using that same data. This is a daunting task, especially considering that millions of new research articles are published each year. Where does one even begin to explore all of that data?
“Each new study contributes more to what we know about a topic, adding nuance and complexity that helps improve our understanding of the natural world. To make sense of this wealth of evidence and get closer to a complete picture of the world, researchers are increasingly turning to systematic review methods as a way to synthesize this information,” says Grames.
“It is important to find all of the relevant information and to not find too much of it,” says Grames. The way to perform this search is through something called a systematic review, which Grames says started in the fields of medicine and public health, where keeping current with research can be a question of life or death.
“In those fields, there is an established system with Medical Subject Headers where articles get tagged with keywords associated with the work, but ecology does not have that.” Other fields of research across the scientific spectrum were in the same boat.
The project sprang out of need. In her own process of reviewing, Grames noted she would miss articles and key terms and was interested in finding out how to identify those missing terms on the first try.
“As we were working on this software, we realized there was a much faster way to do the reviews than how others were doing them,” says Grames, “The traditional way was mostly going through papers and pulling out a term and then reading the rest of the article to identify more terms to use.”
Even with fairly specific key words, Grames notes the average systematic review in her field of conservation biology initially yields about 10,000 research papers for bigger projects. It is important to retrieve relevant information, without also retrieving too much irrelevant information.
“Each year, the amount of data just keeps increasing. There are some systematic reviews that if you look at the amount of time they would have taken just three years ago, they would take about 300 days to perform. If the same reviews were done today, they would take about 350 days because the number of publications just keeps going up and up.”
Grames says it took about a month or two to hash out ideas for the software, then she spent a summer writing and fixing the code. The result is an open-source software package called litsearchr.
How it works, says Grames, is that a user will input their best faith effort of putting together a search into a few databases.
“The keywords should be fairly relevant to retrieve articles that are entered into the algorithm to extract all of the potential keywords, which are then put into a network. The original keywords are at the center of the network and are the most well-connected.”
Grames says the time required to develop a search strategy has been decreased by 90%.
Presented with the most relevant articles, researchers then have significantly fewer papers to parse through manually. This review stage itself is partially automated now too, adds Grames.
Litsearchr is part of a collaborative effort by researchers, called metaverse, where the goal is to link several software packages together so researchers can perform their research from start to finish in the same coding language, R.
“Researchers can develop their systematic reviews, import data, and there is even a package that can write up the results section for the systematic review,” says Grames.
Grames and her team set up the software so that it could be used by anyone, whether they can code or not, using templates ready to upload information to. There is also a detailed step-by-step video to take users through the process.
By keeping the software open source, Grames says debugging and editing is improved because users can point out details that need attention,
“Every time I get an email, it is so exciting. It is nice to have it open because people can let me know when there is a typo.”
The software is currently being used by researchers in scientific fields such as nutritional science and psychology and for a massive undertaking of screening all papers pertaining to insect populations across the globe. Grames says it is nice to have the software in place to be able to take on such a big project. “There is no way we could do this project without the level of automation we get using litsearchr. I built this out of a need from another project, but this software is making it possible to do even bigger analyses than before.”
Go to Source