Risk and NLP: detection, prevention, management

JEP/TALN/RECITAL 2016 workshop

PROGRAMME OF THE WORKSHOP

July 4th, INALCO (65 rue des Grands Moulins, Paris 13), room 4.07 (4th floor)

  • 14h00-14h15: Natalia Grabar et Ludovic Tanguy : Opening of the workshop (paper, slides)
  • 14h15-14h30: Celine Raynal et Christian Blatter : Méthodes d'analyse textuelle pour l'interprétation des REX humains, organisationnels et techniques (slides)
  • 14h30-14h55: Anne Condamines : Vers la définition de nouvelles langues contrôlées pour limiter le "risque langagier" (paper, slides)
  • 14h55-15h20: Nelleke Oostdijk, Ali Hürriyetoğlu, Marco Puts, Piet Daas and Antal van Den Bosch : Information extraction from the social media: a linguistically motivated approach (paper, slides)
  • 15h20-15h45: Natalia Grabar et Thierry Hamon : Exploitation de différentes approches pour détecter et catégoriser le risque chimique et bactériologique (paper, slides)
  • 15h45-16h00: Sandra Cestic et Iris Eshkol-Taravella : Prévention des risques liés à l'environnement de travail : constitution d'un corpus oral en vue de son traitement automatique (paper, slides)
  • 16h00-16h30: coffee break
  • 16h30-16h55: Natalia Grabar et Frantz Thiessard : Types de risque médical et leur traitement avec des méthodes de TAL (paper, slides)
  • 16h55-17h20: Émilie Merdy, Juyeon Kang et Ludovic Tanguy : Identification de termes flous et génériques dans la documentation technique : expérimentation avec l'analyse distributionnelle automatique (paper, slides)
  • 17h20-17h45: Celine Raynal, Assaf Urieli et Eric Hermann : PLUS : pour l'exploration de bases données REX (paper, slides)
  • 17h45-18h15: Closing discussion

DESCRIPTION

He who doesn't risk never gets to drink champagne. (Russian proverb)

The risk, which is a common feature to individuals and society, covers a wide rage of situations and areas. As a matter of fact, several types of activities can meet abnormal situations which consequences may be heavily harmful.

Among the domains, those related to transport (air, road, rail, sea or space), to energy (power plants, oil industry), to biomedicine (medicine, pharmacology) and to heavy industry (metalworking, chemistry, building) are the most concerned by the extent of possible damages: indeed, they are often using surveillance and monitoring systems for the detection, prevention and limiting the risks. Several of these systems exploit the language dimension of the data (accident and incident reports, feedback on interventions, surveillance of the communication flow, etc.): thus, the Natural Language Processing (NLP) applications can provide a contribution for the risk management.

Yet, in some areas, detection of dangerous situations may be based on the analysis of large amounts of textual and communication data, often specific to a given area, such as it happens when analyzing social networks [Neubig et al., 2011] and scientific literature [Hamon et al., 2010].

For some years now, the research trends testifies about special interest for NLP works and applications addressing the risk detection, prevention, and management. Let's mention for instance:

The existing work is mainly done and known within specific communities and the corresponding application areas. The purpose of our workshop is to provide a transversal exchange space between various actors working on detection, description, analysis and prevention of risk, and using for this the NLP resources and methods.

Among the approaches of interest, we can consider non-exhaustively:

  • analysis of feedback reports (reports on accidents and incidents): assistance to experts;
  • detection of potential risk in the existing activity reports: identification of "emerging signals";
  • detection of potential risk in communication: identification of explicit risk in communication between users of a system and/or domain experts;
  • assistance to communication and writing of critical documents: definition and verification of controlled languages and writing conventions;
  • development of resources for the prevention and management of risk: information extraction and knowledge engineering.

The objective of this workshop is to gather a large community from various domains and applications, in order to identify real needs of the ground experts and to discuss on solutions which can be proposed by modern NLP technologies. Hence, the workshop is organized for researchers and industrial actors.

References

[Blanchemanche et al., 2009] Blanchemanche, S., Buche, P., Dibie, J., Feinblatt-Mélèze, E., Ibanescu, M. et Rona-Tas, A. (2009). Ontology building : an application in food risk analysis. In Actes de la 8ème conférence internationale Terminologie et Intelligence Artificielle (TIA), Toulouse.
[Blatter et Raynal, 2014] Blatter, C. et Raynal, C. (2014). Méthodes d'analyse textuelle pour l'interprétation des REX humains, organisationnels et techniques. In Actes du Congrès Lambda Mu 19, Dijon.
[Hamon et al., 2010] Hamon, T., Grana, M., Raggio, V., Grabar, N. et Naya, H. (2010). Identification of relations between risk factors and their pathologies or health conditions by mining scientific literature. Studies in Health Technology and Informatics, 160(Pt 2):964-8.
[Mercantini, 2008] Mercantini, J.-M. (2008). Construction d'ontologies pour la résolution de problèmes de sécurité : une étape vers l'ontologie du risque. In Actes du 16eme Congrès sur la Maîtrise des Risques et la Sûreté de Fonctionnement, Avignon.
[Neubig et al., 2011] Neubig, G., Matsubayashi, Y., Hagiwara, M. et Murakami, K. (2011). Safety Information Mining : What can NLP do in a disaster ? In Proceedings of the 5th Interna- tional Joint Conference on NLP (IJCNLP), volume 11, pages 965-973, Chiang Mai, Thailand.
[Tanguy et al., 2015] Tanguy, L., Tulechki, N., Urieli, A., Hermann, E. et Raynal, C. (2015). Natural language processing for aviation safety reports : from classification to interactive analysis. Computers in industry, In press.

SUBMISSION PROCESS

Two presentation formats are possible:

  • long paper: paper with 6 to 8 pages (TALN format) for a 20 minutes oral presentation, in order to present work with solid results or ongoing work.
  • short paper: abstract with 2 to 4 pages (TALN format) for a 10 minutes oral presentation, in order to present new research problematics, issues or questionings.
Submissions in two languages are acceptable: French, English
LaTeX style-sheet, and LibreOffice and Word models are available on the main conference website (https://jep-taln2016.limsi.fr/styles/jeptaln2016-v2.zip)
Submissions must be done in PDF format through the dedicated website: https://www.easychair.org/conferences/?conf=taln2016