As the birthplace of P2P lending, the UK has been the fertile soil for the development of fintech. In recent years, many outstanding enterprises have emerged. We travelled to London in June 2017 to visit several representative fintech companies, and here we share one of a several special reports on their innovative business models.
Iris.AI was founded in 2015. The company used artificial intelligence to help enterprise and university-based research and development (R&D) to screen academic papers and assist researchers. More than 3000 academic papers are published every day. For researchers, the question of how to find the most relevant information among millions of ever growing academic papers in a short amount of time is a major challenge. Iris.AI wants to use artificial intelligence to help researchers solve this problem.
Different from Keyword: Concept Search Combining with Context
Iris.AI has recently released its version 2.0, which uses fields or abstracts input by the user, and extracts the key concepts that contain the semantic background, to help the user find more relevant topics and other papers. After the user enters the paper link, Iris.AI is able to construct a knowledge map for them in a short period of time.
Iris.AI differs from traditional keyword search tools because it combines the specific meaning of the term in the paragraph, so it is possible to screen for more relevant paper materials. Iris.AI selects the keywords in the text using the TF-IDF (term frequency-inverse document frequency) algorithm, where the text is the abstract part of the paper. Then, Iris.AI combines these keywords with their specific context semantics. Finally, Iris.AI classifies and clusters these contextual concepts. Each of these steps is done through machine learning. So far, Iris.AI has studied about 77 million English-language academic papers in public databases.
Iris.AI’s Target Customers & Customer Acquisition
Iris.AI is mainly aimed at R&D departments in biotechnology, manufacturing (especially the material industry) and engineering, as well as some colleges and universities. Iris.AI focuses on these fields because Iris.AI better able to extract natural science information at this stage. Social science has higher requirements for the semantic understanding and needs more powerful learning ability, so Iris.AI will continue to explore in this area.
The company will train its Iris.AI tools for different customers. After obtaining the customer's database license, Iris can learn the papers in the database and serve specific customers on this basis. To help users better understand and use Iris, the company launched an activity called Scithon (Science Hackathon), that is, a collection of interdisciplinary researchers, grouped to challenge different academic problems to help Iris.Ai build its knowledge graph. Iris.AI’s chief marketing officer, Maria Ritola says that the purpose of holding Scithon’s is to help researchers further understand and experience Iris. Because the search by keyword has been used for such a long time, Maria hopes that researchers can learn the advantages of Iris and incorporate this kind of research activity.
Profit Model
The company profits by providing specific Iris.AI products and solutions for R&D departments in enterprises and universities to build their knowledge profile.
Q&A
Below is our interview with Maria Ritola, CMO and Co-Founder of Iris.AI:
Q: Is it possible to verify that Iris is more efficient than keyword search tools?
A: It can be verified in two ways. First, research in the university shows that building the knowledge graph through artificial screening takes 3 months, but it takes only two days if using Iris; and the confidence level of this test is 90%. Another proof is that in our Scithon, we know that the teams who used Iris were more efficient and used less time.
Q: What's the difference between different disciplines when using Iris?
A: Iris can be better applied in natural sciences now, such as life science, materials science and physics. Iris can also be used in the field of social science, but its understanding of the context and semantics is vaguer, so it is mainly used in the natural sciences.
Q: What's the difference between AI and other artificial intelligence startups?
A: Some similar companies are using article references to find relevant content, but we are not. We think that references tend to have some bias, and many publications need a specific reference article. We still focus on the text itself and look for the most relevant information.
Q: How about the company's future development?
A: We hope that Iris not only become a search tool for the papers, but also be more proactive and become a research assistant for researchers. In terms of technology, Iris.Ai starts by extracting the assumptions in a specific article and then premise of the assumptions, building a hypothesis graph in a specific theme, find the associations and using Iris to make new assumptions. The company is seeking to cooperate with large companies which have strong R&D needs, to further explore customers' needs and improve their products.
Q: Have you considered entering the areas outside academic papers?
A: We can now use Iris to retrieve academic papers in different areas including fintech. Faced with other text such as industrial reports, we can expand the learning scope, but it will take time to let Iris learn the text outside academic papers.
