Thibault joined IRT SystemX in January 2014 to begin a thesis on bandit algorithms for real time data capture in social networks. He about his research in the IMM project (Multimedia Multilingual Integration), under the tutelage of LIP 6 (Université Pierre et Marie Curie).
What is your thesis subject?
My thesis is focused on targeted data collecting in a social media. The purpose of my work was to develop machine learning algorithms that would be capable of automatically extracting knowledge from my data. More especially, the algorithms were aimed at selecting the most relevant users to follow in a social media, automatically because it is impossible to follow everybody (technical constraints and limits to human capacity to track all the activity in a social media). I applied these algorithms to Twitter to identify the most influential people in the network, based on given topics such as Brexit, the American presidential elections or the Rio Olympic games.
What have you gained from your PhD?
It was a powerful, rich experience. I learned autonomy and how to define problems clearly. When you work on a PhD, the difficulty is to keep on course and remain focused on the subject while exploring new avenues of research. SystemX gave me a structured framework and academic and industrial support that enabled me to develop my skills as a young researcher.
What is your best memory at SystemX?
No particular event, but rather a whole period during which it was particularly enjoyable to work in the IMM project. It was when three CRÉE* interns were working in the project. I found the atmosphere extremely pleasant and the exchanges between the various people (engineers, linguists, jurists etc) gave the IMM project huge impetus.
What are your plans for the future?
I have just begun to work at Snips as a Data Scientist. Snips is a startup that makes vocal aid tools enabling interaction with connected objects via speech. I am working on deep learning algorithms that will enable us to detect the various entities within the sentence. This work is very interesting because it requires us to be aware of the latest scientific advances in the field and it offers concrete solutions.
More about Thibault Gisselbrecht
Qualification: Engineering graduate, Supaéro
Thesis subject: Bandit algorithms for real time information collection in the social networks
project: IMM –Integration Multilingual Multimedia
*CRÉE : Cooperation – Research – Student – Enterprise
