Audio content indexing

By creating an index from the spoken content of very large audio databases, we make the content accessible through search engines similar to Google.

In order to accomplish that, we must combine speech recognition with another expertise in indexing and searching, which allows to find the searched terms in spite of recognition errors, pronunciation or spelling variants, and the limited vocabulary of the recognition system.

This expertise has been applied to some projects relating to the cinematographic archives of the National Film Board (NFB) or the collected testimonies of the Bastarache investigative commission.

Related technologies: recognition engine, finite-state transducers.

Teams

Releases

Recent news

  • Valorisation de la recherche québécoise
    16/09/2020

    Le CRIM salue l’importance que le Ministre Pierre Fitzgibbon accorde à la valorisation de la recherche québécoise et l’ampleur des ressources qu’il y consacrera.

    +

Upcoming event

  • Santé et sécurité du travail 2020 - Événement les Affaires
    23 September 2020 8:30
    Présentation en ligne
    Le CRIM est fier d'être partenaire de la 10e édition de la conférence Santé et sécurité du travail organisée par les Événement Les Affaires. Présentation en ligne.
    +

Recent Publications

  • An end-to-end approach for the verification problem: learning the right distance

    +
  • The Indigenous Languages Technology Project at NRC Canada: an empowerment-oriented approach to developing language software

    +