Great results from our first Scithon™!
How to build a rocket with composite materials? Together with the leading European research institute Swerea SICOMP and Chalmers University we organised a science hackathon, or scithon, as we call it. The goal of the 4-hour sprint was to map out solutions to this space challenge and, in the process, get a grasp of where Iris.AI 2.0* stands compared to traditional science discovery tools.
On September 20th two teams of cross-disciplinary Masters and Ph.D. students from fields spanning from mechanical engineering and industrial design to computer science, astrophysics and entrepreneurship were handed a research challenge: Is it possible to make a reusable rocket made completely out of composite materials?
This challenge provided by Swerea SICOMP is particularly difficult due to issues like the performance of composites at extreme temperatures, the limited durability against UV and space radiation, chemical resistance issues with rocket fuels, and oxidation in high concentrations of oxygen.
After introducing the challenge and the rules of the game, the teams were pitted against each other. They both had four hours to achieve two goals: (1) map and categorise related scientific articles; and, (2) summarise the key findings by skimming through the categories and papers. Only one of the teams had access to Iris.AI.
The specific criteria they would be evaluated on were the relevance, breadth and completeness of the research papers identified. Teams’ work was also assessed based on the quality of the conclusions drawn, including elements like issues surfaced, key trends and current research directions identified.
After the sprint, an expert panel evaluated the results obtained by both teams. Team 2, using Iris.AI as the tool, generated a score of 95%. Team 1, using the current market standard product, scored 45%.
The number of generally relevant papers identified was similar for both teams. The different angles covered by these papers (with categories like validation research vs. evaluation research vs. solution proposals vs. philosophical papers vs. opinion papers vs. experience papers) was broadly similar, too.
The scithon jury attributed a significantly higher score to the Team 2, i.e. the team that had used Iris.AI, on three accounts: (1) finding three papers with a top score in terms of fitting the problem statement; (2) showing higher quality of key findings structured around identified topics; and, (3) drawing superior conclusions and summarising the relevant knowledge.
While the team using an existing market standard tool struggled to formulate the relevant keywords to optimise their searches, i.e. facing issues around dated terminology, members of the jury from our co-organisers Swerea SICOMP were particularly impressed by the papers identified by the team using Iris.AI. More specifically, the team using Iris.AI found papers around silicon-based nanoparticles and a distributed health monitoring system for reusable liquid rocket engines. These two key avenues of research could bring us a lot closer to building reusable rockets made of composite materials!
This means that version 2.0 of Iris.AI, with its full text search, unbiased mathematical architecture, neural topic modelling and visual navigation interface features, is beginning to show significant value added for researchers looking to speed up the effective discovery and deployment of scientific research.
The scithon also allowed us to gather invaluable feedback from researchers around the importance of features like filtering (including search criteria refinements) and interaction (including discarding concepts presented by Iris.AI in results maps), which will be included in our near term product roadmap.
The next scithon will be organised on October 28th in Stockholm in collaboration with Iris.AI and Future Earth. If you are in the area and would like to join us to identify solutions to climate change, contact Maria at email@example.com.
*The new version of Iris.AI will be launched on the 22nd of September. Be the first one to hear about it by signing up to our newsletter at www.iris.ai.
Interested in having a look at the Scithon material? Here’s the Dropbox link to view the results delivered by both teams as well as the full version of the problem statement.