Page en français

Camille GUINAUDEAU

Contact

INRIA - Campus de Beaulieu
35 042 Rennes CEDEX
FRANCE

tel : +33 2 99 84 25 93
fax : +33 2 99 84 71 71
e-mail Camille.Guinaudeau [at] irisa.fr

Assistant: Elodie Lequoc, +33 2 99 84 74 37, Elodie.Lequoc [at] inria.fr

PhD
Teaching assitant at University of Rennes 1

Research interests

  1. TV Stream Structuring
  2. Natural Language Processing
  3. Information Retrieval

Current research activities

  • Automatic structuring and enrichment of TV streams
  • PhD work
  • Advisors: Pr. Pascale Sébillot and Guillaume Gravier
  • The goal of my PhD is to develop automatic techniques based on speech to enable fine structuring of TV stream. The structuring task is divided into two main steps: topic segmentation and characterization. The topic segmentation step aims to segment TV programs extracted from the stream into thematically coherent TV segments (news reports for example). The characterization step serves to inform users about the content of (segments of) TV programs. It is also useful to link together topic segments that adress the same story. To accomplish this structuring goal, speech is accessed by means of automatic transcription of the speech contained in TV programs. Then NLP methods are applied on these transcripts. These methods are mainly based on lexical information but additionnal clues specific to spoken documents (such as confidences measures associated with transcripts and acoustic clues) are used to adapt NLP techniques to speech specificities (transcription errors, lack of vocabulary repetition).

  • Using prosodic information for automatic structuring of TV programs
  • "Spoken Language Processing Group" Affiliate at Columbia University, New York, NY, USA (June-September 2010)
  • Advisor: Pr. Julia Hirschberg
  • In spoken documents, the speaker's emphasis on a particular word is an important clue concerning the relevance of a word in the document. In order to take into account the speaker's intention, prosodic information was extracted from TV programs and was associated with lexical information in order to improve the structuring techniques developed during my PhD.

    • Publications

    Complete list

    Technical skills

  • Programming languages: Perl, JAVA, C/C++, Prolog, SCHEME, SQL
  • Operating systems: Linux, Windows, MacOS
  • Methods: Statistical and probabilistic methods for Natural Language Processing (NLP), Information Retrieval, Data Mining.

    • Teaching

  • 2011-2012: Teaching assistant at University of Rennes 1 - 96 hours
  • Introduction to algorithmics with SCHEME (B. Sc. level)
  • Introduction to programming with JAVA (B. Sc level)
  • Operating systems and networks (B. Sc level)
  • UML (B. Sc level)
  • Audiovisual streams indexing (M. Sc level)
  • Information retrieval in textual documents (M. Sc level)

  • 2010-2011: Teaching assistant at University of Rennes 1 - 10 hours
  • Information retrieval in textual documents (M. Sc level)

  • 2009-2010: Teaching assistant at INSA, Rennes - 80 hours
  • Databases (B. Sc level) - 30 hours
  • Introduction to programming with language C (B. Sc level) - 14 hours
  • Introduction to object-oriented programming with JAVA (B. Sc level) - 36 hours

  • 2008-2009: Teaching assistant at INSA, Rennes - 47 hours
  • Databases (B.Sc. level) - 28 hours
  • Introduction to algorithmics with SCHEME (B. Sc. level) - 9 hours
  • Advanced programming with JAVA (B. Sc level) - 10 hours