Natural language processing
Data consolidation and natural language processing used in the fight against radicalisation
Date:
Changed on 06/05/2021
In support of the fight against radicalisation and to prevent potential terrorist attacks, CounteR, a multi-partner EU project, is aiming to use data consolidation in order to develop a platform that will allow police forces to detect individuals being trained to commit violent acts. Let’s take a closer look.
CounteR has its roots in an initiative put forward by Frédérique Segond, director of Inria’s Defence and Security mission, which was set up to coordinate research carried out by the institute that is applicable to French and European sovereignty. ALMAnaCH, a research team at Inria Paris which focuses on artificial intelligence through a combination of natural language processing (NLP) and digital humanities, was a logical choice to join the consortium as NLP experts.
Strength in numbers: a collaborative project with a range of international partners
CounteR boasts specialists in information extraction and data mining, in addition to experts in the psychology of radicalisation and institutional services such as France’s Ministry for the Interior (see the list of partners). Because of the fundamentally multilingual nature of the project, full control over all aspects of the processing chain is required, from complex flow data extraction to multimodal and decision-support semantic processing.
Djamé Seddah, head of the project for ALMAnaCH at Inria Paris, has been keen to stress how important these partners are:
I don’t think a team working on its own would be capable of deploying this sort of chain for multiple languages and data sources in such a short space of time.
The European dimension of the project is very important to Djamé Seddah, head of the CounteR project for ALMAnaCH. It is through the EU that he has enjoyed total freedom of movement, allowing him to do postdocs in Luxembourg or in Ireland as easily as if they had been in Bordeaux or Nice. This enabled him to build a professional network which has lasted far beyond his initial collaborative endeavours, significantly influencing his scientific and personal decisions.
Cutting edge natural language processing technology used to tackle radicalisation
As a partner in the project, Inria’s aim is to provide institutional services with “proofs of concept”: cutting-edge natural language processing technology that will allow them to anticipate the formation of critical zones. This technology will be combined with psychological expertise on the mechanisms of radicalisation in order to develop a comprehensive solution that will help police forces to understand the “when”, the “where” and the “why” of this growing phenomenon.
Verbatim
Exploring sources of noisy or multimodal data is a priority for us as we seek to put our research into practice.
Auteur
Poste
Head of the project for the ALMAnaCH team at Inria Paris
The primary objective for ALMAnaCH is to further develop existing textual analysis and information extraction technology so that it can be used in situations where data flows are not only noisy, but also multilingual and multimodal. The researchers are also keen to make sure their technology can be used in sensitive contexts where the robustness and quality of the analysis are key parameters. There is much anticipation surrounding the first series of experiments on real-life data.
At the same time, the challenge is for this monitoring platform to be used for what it is: the role of CounteR is to consolidate data from diverse sources into an analysis and early warning platform for exploring data and predicting potential risks. The aim is for this to be used as a frontline tool in community policing, investigating communities and connected risk factors as opposed to arbitrary targeting and surveillance.
Increased opportunities through EU funding
Having access to EU funding, not just national funding, is vital when it comes to scaling up. Being able to process this type of information in real (or near-real) time on such a large scale, while deploying increasingly complex technology, requires different skillsets and bigger budgets.
Verbatim
Experience acquired in the context of previous ANR (the French National Research Agency) projects that we’ve undertaken (Sequoia, Edylex, Passage, SoSweet, ParSiTi, etc.) within Inria means we are capable of taking on projects of this size.
Auteur
Poste
Head of the project for the ALMAnaCH team at Inria Paris
The CounteR project is funded under the third pillar of the former framework programme H2020. This category is dedicated to the societal challenges that the EU identified as currently facing and of importance for its Member States. The CounteR consortium applied to the following call SU-FCT02-2018-2019-2020 — Technologies to enhance the fight against crime and terrorism.
A project the size of CounteR could only have been rolled out in the context of an EU collaboration. Collaborative EU research projects ensure a good level of granularity, a balance between administrative constraints linked to funding amounts and almost total scientific freedom with regard to partners. Regular reports on the progress of the project, which are used to fine tune expectations between scientific and technical findings and the reality of the data, are particularly important.