Making it easier for researchers and entrepreneurs to seek out funding
Arrandi. Although my name might sound Mediterranean, it is in fact a Gallicised phonetic pronunciation of the English term “R&D”. This multilingual play on words makes my methods clear from the outset: text mining through lexical exploration. My aim is simple: to provide researchers and entrepreneurs with a quick and effective way of sifting through the mass of calls for proposals put forward at a regional, national, European and international level by all manner of different financial backers to identify those which best correspond to their needs and expectations.
The effort involved in putting together an R&D project can be overwhelming. First you have to identify suitable sources of funding, before then seeking out and deciphering documents which can be anywhere from a few dozen to several hundred pages long, and finally putting together an application. “The process is so taxing, relative to its success rate, not to mention how time-consuming it is, that some researchers give up”, explains Pierre Séry, the man who created me. “A 2017 Inserm study revealed that it took on average 23 person-days to put together a basic R&D project, the equivalent of more than €10,000. For more complex projects this can get up to 270 person-days, more than €120,000.” I have the capacity to reduce the time spent by researchers and companies seeking out funding for future projects by at least 50% by automating the data collection and evaluation process.
Natural language processing and data visualisation
I come in the form of a platform that is accessible via subscription, where users are asked to enter a number of search criteria. Using artificial intelligence (AI), I then extract the most relevant calls for proposals from the tens of thousands of pages of them found online based on the information provided (such as the type of partner required, scientific field and budget) before presenting them in an easy-to-view format. For this I employ two tools: natural language processing (NLP) for data selection, and data visualisation for presenting the results. “I have quite a visual way of thinking, which mind mapping helps me to present in a clear and accurate way”, explains Pierre Séry. “The aim was to take inspiration from these visualisation methods in order to present the key information on calls for proposals, thus making them easier to assess. This information is analysed in advance using NLP, which has evolved considerably over the past five years, thanks in large part to improvements made to search engines.” That said, search engines operate chiefly based on a referencing system involving the use of keywords, creating bias and potentially limiting the scope of any search. I am not like that. Although the data scientist Alexis Frisson, an associate of Pierre Séry, used existing tools in my development, he adapted them to my requirements. One of the adaptations he made was to incorporate the specific terminology of the jargon used in calls for proposals.
Making searching for calls for proposals less time-consuming and more efficient
I am currently in the process of being finalised, as part of which 200 potential users have been carrying out tests aimed at further improving me prior to my official launch, scheduled for 1st April 2023. These initial tests have not just been promising, but have confirmed both my capacities and how useful I am: a number of use cases have shown that I am capable of identifying a suitable call for proposals in just a few minutes, where a manual search would have taken several hours or even several days. Here’s a testimony from Pierre Séry: “When we joined Inria Startup Studio last April I spoke with a Professor from Loria who was very interested in our work - since 2019 he had been working on a project that he hadn’t managed to get funding for, despite his best efforts. Using Arrandi, it took us just half an hour to identify a call for proposals, which he applied to in June. Last week we were noticed his application had been approved and was going to receive close to €300,000 worth of funding.”
Calls for proposals are hundreds of pages long and are written in technocratic language, and it is routinely a struggle to find the best projects for you from both a scientific and a financial perspective. Arrandi summarises them using a few key words, presenting them in a clear way. It’s a bit like a pitch for a film. It saves researchers a considerable amount of time and energy, and it’s also efficient: through it we were able to identify ASTRID, a call for proposals financed by the French Defence Innovation Agency and implemented by the ANR (the French National Research Agency), which we were selected for.
An adaptable methodology and tools
The results from my validation tests have raised hopes, and I have the ambition to meet them. My target is to attain 100% growth each year over my first three years, reaching turnover of one million euros at some point in my fourth year. With that in mind, I’m already thinking ahead to the future and how I can draw on my current expertise in order to tackle other markets. “In a number of sectors searching for information is complicated by the quantity and the complexity of texts. This is a problem faced by legal startups, for instance, who have to process vast quantities of legislative and regulatory documents in order to give their clients the best advice. Our aim is to further develop the tools and methodology we currently have for use in other specific areas.” The future looks promising. I have already been contacted by a leading French construction company, asking me if I would be capable of adapting my platform to their specific needs. My answer was in the affirmative: I have both the means and the desire.
Inria Startup Studio: “A remarkable working environment”
Familiar with the challenges and constraints of calls for proposals, the two men behind Arrandi developed innovative processing methodologies. Pierre Séry has solid, practical knowledge of the subject: “I have fifteen years of experience research and innovation funding, during which I occupied a number of roles. When I was working at the CNRS I had to seek out opportunities of funding, and then when I was working for local authorities such as Grand Nancy or the Grand Est region it was my job to offer it.” Alexis Frisson, meanwhile, drew on his skills as an AI engineer and specialist in natural language processing to develop an innovative system for mapping and analysing existing offers. All that was left was to find the best possible environment for them to work in. “Inria Startup Studio offered a remarkable working environment. It paid our salaries for a year, allocated us an operating budget for building our project, and put us alongside researchers and teacher-researchers on a daily basis.”