Section: Application Domains

Automatic and semi-automatic spelling correction in an industrial setting

Participants : Kata Gábor, Pierre Magistry, Benoît Sagot, Éric Villemonte de La Clergerie.

NLP tools and resources used for spelling correction, such as large n-gram collections, POS taggers and finite-state machinery are now mature and precise. In industrial setting such as post-processing after large-scale OCR, these tools and resources should enable spelling correction tools to work on a much larger scale and with a much better precision than what can be found in different contexts with different constraints (e.g., in text editors). Moreover, such industrial contexts allow for a non-costly manual intervention, in case one is able to identify the most uncertain corrections. Alpage is working within the “Investissements d'avenir” project PACTE, headed by Numen, a company specialized in text digitalization, and three other partners. Kata Gábor and Pierre Magistry have worked as PACTE-funded post-docs until the end of the project in March 2015.