The Artificial Intelligence Platform, 2019 edition (PFIA2019) will take place in Toulouse in early July. The annual symposium’s objective is to gather researchers, companies and students for conferences and workshops focused on Artificial Intelligence (AI).
This year’s edition will take place in the campus of Toulouse 1 Capitole University. It is organized by IRIT, the Toulouse Institute of Computer Science Research, and AFIA, the French AI Society, with the support of many other institutional and industrial sponsors. The PFIA2019 consists of eleven conferences or challenges, including TALN, the French conference on Natural Language Processing (NLP).
As a prequel to TALN, a “hackaTAL” will take place (from the words “hackathon” and TAL, which is the French acronym for NLP). This hackaTAL has two challenges:
- topic and argument extraction from the written results of the Grand Débat national survey launched in early 2019 in response to the widespread gilet jaune protests.
- chatbot creation for legal problem solving
There will be about 40 competitors (mostly students in the NLP/AI field) who will work in small teams and to apply state-of-the-art techniques to solve one of these two challenges, using real-life data.
YSEOP will sponsor the first challenge, while the second is sponsored by a French NLP and chatbot company, Synapse (located in Toulouse). The challenge is organized by Damien Nouvel, from Paris Sorbonne University (INALCO). Click here to learn more about the HackaTAL
Our interest in this challenge
One of YSEOP Lab’s goals is to gain visibility in the area of cutting edge research, which is well supported by our involvement in the hackaTAL event.
Why did we choose this specific challenge about argument extraction in the Grand Débat? The Grand Débat was an opportunity for French people to express their thoughts about the current state of their country and what could be better in response to the widespread gilet jaune or yellow jacket protests. If it was one of Europe’s most significant exercises of democratic consultation with over 2 million online contributions, 10,000 local meetings and 16000 complaint books as well as a series of citizen assemblies. The results are a huge body of work (more than 250,000 text contributions) that was released as an open-source data. It will certainly be a subject for research for several years (the French Research Agency has raised a call for projects on this corpus). For text-mining/NLU research, it is an invaluable resource.
The other motivation for our involvement relates more to possible product evolutions, which may lead to a new Lab research product. While YSEOP’s Augmented Analyst is perfectly able to draw conclusions from structured data, there is a lot more work to be done in interpreting unstructured data which makes up on average 80% of the information within a company. While Augmented Analyst can currently interpret a lot of information from structured data. being able to understand unstructured data from company documents could add an additional dimension to report generation.
Automatic identification of reasons (causatives) is a part of the so-called “argument-mining” field in NLP. We will be able to test state-of-the-art techniques applied to real-life corpus (even from very different fields) and share ideas with researchers during these two HackaTAL days. This is also an important part of the Lab’s activity.