As part of my teaching responsibilities, I coordinate the following courses at CentraleSupélec:

• Information systems and programming, 1st year (with Guillaume Mainbourg).
• Cloud and distributed computing, 2nd year.
• Introduction to machine learning, 3rd year (with Myriam Tami).
• Big data algorihtms, techniques and platforms, 3rd year and Master Artificial Intelligence.
• Introduction to databases, Master in Data Sciences & Business Analytics - Essec & CentraleSupélec.

I give lectures and supervise lab assignments and tutorials in the following course at CentraleSupélec:

• Computer networks and security, 1st year.

I also coordinate a Big Data course with Stephane Vialle at Polytech Paris-Saclay and Ecole Centrale Marseille.

My postdoctoral research, coordinated by professor Chantal Reynaud, was funded by the research project DataBridges: Data Integration for Digital Cities (ANR 11 EITS 003 05).

The goal of DataBridges was to create generic software platforms to enable applications that integrate, compare, query and deploy complex, semantically enriched city data.

In this context, my research focused on the study of efficient methods to identify, extract and semantically annotate data contained in tables, such as Excel files or HTML tables. In fact, tables are a very valuable source of data concerning cities and, in particular, they very often contain information, such as statistics, which are very difficult to identify in unstructured web pages.

As part of my research, I designed an algorithm for identifying and annotating named entities in Google Fusion Tables (Google Fusion Tables has since been discontinued). Unlike other entity annotation algorithms, which can identify entities in tables only if they are present in knowledge bases (such as DBpedia), my algorithm learns to search the Web for information about unknown entities and use this information to annotate them.

At the same time, as Google Fusion Tables are shared by Internet users around the world, it is not uncommon to find tables where the content is written in languages other than English. In order to manage the multilingualism of Google Fusion Tables, I turned my attention to the way Wikipedia links, through interlanguage links, articles that talk about the same subject in different languages. Wikipedia’s cross-language links can be used to easily translate concepts extracted from Google Fusion Tables; unfortunately, there are many Wikipedia articles that do not have cross-language links to the corresponding articles. Therefore, I proposed an algorithm that automatically identifies with very good accuracy the missing interlanguage links in Wikipedia.

As part of my postdoctoral research at the University of Maryland, coordinated by professor Hanan Samet, I participated in the development of
NewsStand, a system that automatically aggregates online news from multiple websites and present them on a map. The map allows Internet users to easily access news about the geographic locations they are interested in.

The main theme of my research was geocoding, which is the process of identifying and disambiguating place names (references to geographical localities) in a text document. In fact, one toponym (e.g. Paris) can identify several geographical localities (e.g. Paris, France or Paris, Texas). I have described an algorithm for geocoding toponyms based on the observation that newspaper articles are usually addressed to people living in a specific locality or region and, therefore, often make references to entities relevant to that locality or region. Therefore, each occurrence of Nôtre-Dame in articles in Le Parisien should be interpreted as a reference to the cathedral in Paris rather than to the cathedral in Strasbourg, unless the article contains explicit indications that lead to a different interpretation.

My research activity, supervised by professor Bruce Reed and Michel Syska, was carried out in connection with my doctoral research.

The first part of my thesis focused on an algorithm that creates the rectangular dual of a planar graph. One step of the algorithm consisted in eliminating the separating triangles in a graph. My work at Inria - Sophia Antipolis was aimed at studying this problem.

The main theme of my doctoral research, supervised by professor Massimo Ancona, was the visualization of large dense graphs. A dense graph normally contains many edges whose visualization creates a “visual clutter” that makes it difficult to read the graph and, consequently, to interpret the data represented by the graph.

My thesis describes algorithms to visualize a graph in order to reduce the visual clutter through its rectangular dual. For this purpose, it is necessary to create the rectangular dual of the graph. The rectangular dual of a graph is a subdivision of a rectangle into as many rectangles as the nodes of the graph, with the constraint that two rectangles are adjacent if and only if the two corresponding nodes are adjacent in the graph. In order to construct the rectangular dual of a graph, the graph must meet a set of conditions. Therefore, a large part of my research has focused on (and resulted in) formalizing algorithms that modify the graph to meet these eligibility conditions.

As far as visualization is concerned, one of the main results of my thesis is the description of an algorithm which creates a confluent drawing of a graph using the rectangular dual. In a confluent drawing the intersecting edges of a graph are joined together in a single bundle, which allows to significantly reduce the number of segments and curves in the drawing.

My research was funded by the European project Agamemnon (IST-508013-STP) and supervised by professor Massimo Ancona. Agamemnon aimed at improving visits to archaeological sites through technology. A client-server system was developed to give visitors to two archaeological sites (Paestum, Italy and Mycenae, Greece) an advanced electronic tourist guide. The guide is a software that, by communicating with a server, displays on the visitor’s 3G mobile phone all the information on the monuments visited. On the server side, an image recognition system receives the photographs taken by the visitors’ phones, recognizes the monuments represented in the photos and sends all the information back to the visitor who requested it. I collaborated in the evaluation of the image recognition system and in the preparation of a prototype as a support for tourist visits in a city.

As part of my master’s thesis, supervised by professor Massimo Ancona, I participated in the implementation of a software, called WtX, which allows to write long texts quickly and comfortably on Personal Digital Assistants (PDAs), which are very small to have a keyboard as in the case of a computer. WtX provides a virtual keyboard, a handwriting recognition system and a word prediction system that, with the help of a dictionary, predicts the words that the user is writing. WtX includes a system, which I designed and implemented, that allows users to write whole words using abbreviations. As a result, users can write long texts quickly and comfortably.