Maria Georgescul

Contact information: ISSCO/TIM/ETI, University of Geneva 40 bd. du Pont-d'Arve CH-1211 Geneva 4 Switzlerand

Tel: +41 22 379 8683 Fax: +41 22 379 8689 Office number: 6336

Background and Research Interests

I have been working as a scientific researcher at ISSCO / TIM, ETI, University of Geneva since May 2002. I was previously a research assistant in LITH department at EPFL. I received both a MSc and BSc in Computer Science from Faculty of Computer Science, Al. I. Cuza University of Iasi.

Working in the framework of the Swiss National Center of Competence in Research on Interactive Multimodal Information Management (IM2), I've completed my PhD thesis in April 2007. The thesis is dealing with machine learning techniques for discourse/dialogue segmentation into thematic episodes.

My research interests are in

• designing and empirically evaluating machine learning techniques for natural language processing tasks
• implementing dialogue systems
• exploiting corpora for translation
• machine translation

Current and past research projects that I've worked on:

• CoNLL-2010 Shared Task - Learning to detect hedges and their scope in natural language text.
• CALL-SLT - Computer-Assisted Language Learning based on speech recognition and machine translation.
• MedSLT - Spoken language translation of dialogues in the medical domain.
• IM2 - Interactive Multimodal Information Management

Research project proposals that I've worked on:

• TEACH - Translating Educational Articles for CHildren. [ Abstract ]
• ACCEPT - Automated Community Content Editing PorTal. [ Abstract ]

Teaching Activities

• First semester: teaching the "Documents structurés multilingues (XML)" course on multilingual structured documents using XML technologies.
• First semester: teaching assistant for the "Technologie de l'information et de la communication multilingue" course.
• Second semester: teaching assistant for the "Exploitation et gestion de données textuelles (corpus)" course on exploiting corpora for translation.

Publications

• A Hedgehop Over a Max-Margin Framework Using Hedge Cues, Maria Georgescul. "In Proceedings of the 14th International Conference on Computational Natural Language Learning (CoNLL)", Uppsala, Sweden, July, 2010.
• Spoken Language Understanding via Supervised Learning and Linguistically Motivated Features, Maria Georgescul, Manny Rayner and Pierrette Bouillon. In Proceedings of the 15th International Conference on Applications of Natural Language to Information Systems, Cardiff, UK, June 2010.

Tools that I've worked on:

• CALL-SLT - a toolkit for Computer-Assisted Language Learning (CALL) which uses speech recognition and speech translation resources to build conversation partners for beginning language students, based on the idea of a "translation game".
• REGULUS - a toolkit for building spoken dialogue systems.
• Java GUI of a speech-enabled application for accessing a meeting database (see also [3]).
• MedSLT - an application for medical speech translation.
• Java implementation of a few evaluation measures, i.e. word error rate and evaluation metrics for topic segmentation: P_k, WindowDiff and Pr_error (see also [7]).
• Browsing a meeting recording in Transcript-based Query and Browsing Interface (see also [8])