Defense of the HDR entitled ‘Natural Language Processing: Studies and Contributions at the Frontiers of Interdisciplinarity’, on Tuesday, December 8, 2020, at 2:00 PM in the Thesis Room of Avignon University (Hannah Arendt Campus – City Center). The defense committee will be composed of:
This dataset consists of 3 TV series with manual annotations: All three files are in .json format and contain TV Series annotated data. Each TV Series is defined by its name, A TV Series contains seasons, defined by their ids. Every season is made of episodes, defined by their ids, titles, duration and fps. Each episode contains two basic kinds of data: scenes and speech segments. Scenes are defined by starting points and are made of shots (Seasons 1 only).A shot is defined by starting and ending positions, and recurring shot ids. The speech segments are defined by their starting and ending points; textual content (here encrypted for copyright reasons); speaker; possible interlocutors.
This dataset contains conversations between Wikipedia editors, which are annotated in terms of various types of abuse, at the level of messages. It aligns two existing corpora:
These scripts aim at analyzing a historical dataset describing the relationships between Roman emperor Trajan and his entourage. It does the following:
These scripts were designed to analyze the European Parliament votes through a multiplex network-based approach. Our tool was applied to data representing the activity of the members of the European Parliament (MEPs) during the 7th term (from June 2009 to June 2014). The raw data describing this activity were retrieved from the It’s Your Parliament website. There were some minor issues with these data, which we had to correct: some MEPs were represented twice, some profiles were incomplete, the policy domains were not defined for all vote texts, etc. These cleaned data, as well as our figures and results, are available on Zenodo.
Thesis defense of Titouan Parcollet, entitled “Artificial Neural Networks Based on Quaternion Algebra,” will take place on Tuesday, December 3, 2019, at 2:30 PM in the Blaise Pascal amphitheater (CERI). The thesis will be presented before a jury composed of: The defense will be conducted in French. You are also invited to the reception following the defense in Room 5. Abstract: In recent years, deep learning has become the preferred approach for developing modern artificial intelligence (AI). The significant increase in computing power, along with the ever-growing amount of available data, has made deep neural networks the most efficient solution for solving complex problems. However, accurately representing the multidimensionality of real-world data remains a major challenge for artificial neural architectures. To address this challenge, neural networks based on complex and hypercomplex number algebras have been developed. Thus, the multidimensionality of data is integrated into neurons, which are now complex and hypercomplex components of the model. In particular, quaternion neural networks (QNNs) have been proposed to process three-dimensional and four-dimensional data, based on quaternions representing rotations in our three-dimensional space. Unfortunately, unlike complex-valued neural networks, which are now accepted as an alternative to real-valued neural networks, QNNs suffer from several limitations, Plus d'infos
On the next 26th of November at 4 PM in the thesis room (Hannah Arendt campus). This HDR entitled ‘Neural Networks for Natural Language Processing’ will be presented before a jury composed of: Reviewers: Mrs. Dilek Z. HAKKANI-TÜR Senior Principal Scientist, Alexa AI, USA Mr. Patrice BELLOT Professor, AMU Polytech’, LIS, Marseille Mr. Frédéric ALEXANDRE Research Director INRIA, Bordeaux Examiners: Mr. Yannick ESTÈVE Professor, AU, LIA, Avignon Mr. Frédéric BÉCHET Professor, AMU, LIS, Marseille
DeCoMaP: Detecting Corruption in Public Procurement The societal benefits of opening up public data are expected to be huge. This is particularly true with Public Procurement Data which are supposed to help discover and dismantle corrupt activities by facilitating critical information, tools, and mechanism for judicial enforcement. In a multidisciplinary project, bridging computer science, economics and law, DeCoMap is intended to collect, process and analyze French procurement data in order to create a software tool for automatic identification of corruption and fraud in public procurement (automated red flagging) and provide normative analytical grid by highlighting the main factors that public authorities should identify and pay attention to. Supported by Transparency International France and Open Contracting Partnership, DeCoMap brings together academic researchers from 7 universities, with strong expertise in procurement and digital law, procurement economics and econometrics, law and economics, graph optimization and complex network analysis. 4 members of Datactivist, a cooperative company that assists organizations from the public, private and non-profit sectors in producing and re-using Open Data, with strong expertise with open data of public procurement, open contracting and open government, complement the consortium. Date: 2019–2024 Website: https://decomap.univ-avignon.fr ANR page: https://anr.fr/Projet-ANR-19-CE38-0004
These scripts were designed to randomly generate signed graphs possessing some form of community structure, in order to assess partitioning algorithms. Various aspects of the graphs can be specified by the user.
This software takes the name of a public person and a period, and retrieve all events available online involving this person during this period. It first performs a Web search using various engines, then retrieves the corresponding Web pages, performs NER (named entity recognition), uses these entities to cluster the articles, and considers each cluster as the description of a specific event. It is designed to handle Web pages in French, but should work also for English.
Laboratoire Informatique d'Avignon — Avignon Université