AI Ireland finds out about SciencePOD’s latest productivity tool, ScioWire
As SciencePOD is launching a new productivity tool, ScioWire, the company’s CEO and Founder Sabine Louët chatted with Mark Kelly, founder of AI Ireland and the AI awards on his AI Ireland Podcast. The mission of AI Ireland is to accelerate the use of AI for the benefit of our society. Here, SciencePOD CEO tells us one of the ways that natural language processing algorithms included in SciencePOD’s newsfeed, based on AI Summarisation, can help knowledge economy professionals, including scientists. This solution helps them gain productivity when writing their own research publications by building on the latest scientific studies on record to provide robust arguments.
The beginning of SciencePOD
What motivated SciencePOD’s founder is simple. “When I moved to Ireland … I couldn’t find much in the way of publishing. So, I felt that there was a bit of a missed opportunity in Ireland,” she says, referring to the fact that Ireland is a nation of highly educated native English speakers. This gave her the idea to specialise in the creation of summarised digital content, used as teaser material to attract the attention of professionals to the latest research findings and discoveries.
SciencePOD and AI summarisation
Explaining how the business came about using AI, she describes the exciting collaboration SciencePOD entered to automate the creation of summaries of the latest research studies to be used as teaser material and raise their profile on a large scale.
The need for new AI summarisation solutions has been sparked by a change. “The way science and research is being reported is changing,” she explains, “Traditionally, we would have a scientific study which has been peer-reviewed, but as time passes by, more and more people do what is known as preprint, so they publish before someone has a chance to review it.” As a result, “the way scientists consume the content, the actual work from their colleague, is also changing.”
She felt we now need to adapt existing solutions. Speed and productivity are very important factors in this new way of working, as scientists create their own research papers and reviews, based on previous work. “This means that we need to adapt solutions to allow the scientists and people working in that knowledge economy to access the research very quickly,” she points out. “So, this is why we turned towards AI summarisation solutions that allow us to save time. It’s that simple,” she adds.
As the field is maturing, Louët also outlines that there are many AI summarisation solutions available. However, many are not bespoke or adapted to the job of summarising research studies.
To tackle this issue, SciencePOD worked in collaboration with a world expert in summarisation, Juan Manuel Torres Moreno, who is a senior scientist at the Avignon Computer Laboratory LIA, at Avignon Université. He has built up over 25 years of expertise through working on summarisation technology and algorithms. “With his help, we were able to create a bespoke summariser that specifically works with scientific studies and is specifically designed to give context to the reader so that anyone who is not an expert can still understand, “ she explains.
Specifically, SciencePOD has used a method called extractive summarisation, which is a more effective approach than machine learning. “To extract the most meaningful part of the original source document, we adopted an extractive approach which is essentially telling the algorithms how to extract the most significant part of the original source document and then presenting it in a way that’s easy to digest for the reader,” Louët says.
SciencePOD also uses statistical methods to determine, which sentences make the final cut. “It’s basically natural language processing and we have statistical methods to give weight to every single word and every single sentence in the paper,” she explains, “And based on that, we do calculations”. Their method also includes sentence compression and other ways of improving the resulting summaries. This method has been tested on thousands of publications which have shown that summaries are sufficiently precise for the purpose of saving time when deciding which paper to read.
A productivity tool
Although it is not meant to substitute reading the full paper, ScioWire can act as a perfect productivity tool for those who cannot review all the studies published at one time:“[ScioWire] is by no means meant to substitute the full reading of the full paper” Louët points out, adding: “It’s used mainly for people who may need to review 100 studies at one time. They don’t have time to read all of them”.
It also means that they no longer have to solely rely on the abstract. “The abstract, which is the natural summary that all studies have, very often is a bit biased because the author might want to apply for the next grant, and they might put a lot of emphasis on one aspect and not on the rest,” Louët explains. By contrast, SciencePOD’s AI summarisation tool provides a more objective reflection of each paper.
Applications of AI summarisation
SciencePOD’s CEO Also outlines the actual applications that stem from such an approach. “We are actually developing a news feed based on these summaries,” she says. Namely, SciencePOD has just launched ScioWire. “We’re extracting automatically all of the open-access studies we have permission to use, and we are summarising them into a news feed,” She says. “So, one of the concrete applications is for monitoring the latest research in a given field.”
The customised newsfeed means that users can tailor their search results to their preferences. The newsfeed also allows users to combine search results such as time, where it’s published, the topic it covers and so on.
This interview was first published in AI Ireland blog. See the original post here