IrisAIIris.ai, a start-up headquartered in Norway, specialized in AI tools and applications to allow humans to make sense of and use all scientific knowledge the world has to offer; hence, it is particularly suitable for interdisciplinary projects where the combination of knowledge from across a range of research fields will be vital to the project’s success.

So far, Iris.ai developed two tools, the Explore and the Focus Tool, which use artificial intelligence to make literature searches more effective and faster. As the name suggests, the Explore Tool is designed to find relevant literature based on free text input or an article, without being limited to keywords. The results are then presented in a visual map of concepts that make navigation easier. The Focus Tool then helps to refine and shorten a reading list of articles, which happens without the user having to read an article, but by including and excluding concepts and topics that appear in the articles, thereby saving a lot of time.


As you can imagine, market researchers and the media could benefit from such an AI-based research tool enormously too.


Iris’ AI works with content-based recommendations: it requires a text of 100 to 500 words in which the question is explained in more detail, for example with background knowledge or the explanation of specific expressions, to be able to find articles that correspond to this question.


All in all, the system accesses around 160 million publicly available scientific papers, and so far, it seems as if the health-, natural-, and computer sciences as well as the engineering sector benefit from it. The reason for that is most probably that the algorithms heavily depend on how much research is accessible in the open access format, but there are by now success stories from the humanities too.


Iris.ai’s approach to artificial intelligence is in the field of Natural Language Processing (NLP) and uses a combination of keyword extraction, word embedding, neural topic modeling and word weighting, based on the similarity of document metrics and hierarchical topic modeling. It can thus understand text input, create a “fingerprint” of the text (based on its concepts and topics), and compare it with the texts in the database. In doing so, it suggests contextually similar articles and can help refine the reading list.


Iris.ai users are mainly students, academic staff from universities, researchers at institutes, employees from research and development departments, or simply science enthusiasts who want to explore the spectrum of open science literature. It short, Iris makes sense for anyone looking to get more scientific material on a topic; hence, media companies could be interested in using Iris for their research too. Iris seems to be simply a great support for anyone in research who must deal with a huge number of publications.


In fact, in times of COVID deniers, Iris’s work could be more important than ever, as they are supporters of the open science idea and want to create opportunities so that even “non-scientists” can find research findings more easily. They even decided to integrate the COVID-19 dataset, CORD-19, and to give all COVID-19 researchers free access to the Iris Premium tool.


In any case, you can expect Iris.ai to continue adding exciting features in their next versions and to continue its road to success.


By Daniela La Marca