Suchergebnisse

Historical documents have specific properties
which make life hard for traditional information retrieval
techniques. The missing notion of orthography and a general
high degree of variation in the phonetic-graphemic representation,
as well as in derivational morphology obstruct the
possibility to find documents upon the entry of a modern
word as the search term. The following paper gives an
overview of existing string approximation technologies as
used in bioinformatics, but also of phonetic approximation
algorithms. It proposes an architecture of combining both
notions, while using Jörg Michael' phonet program to deduct
from graphemes to a phonetic representation and a
levenshtein automaton to allow for fast approximative
matching. The final part of the paper evaluates the suitability
of the approach, while using the levenshtein algorithm in
its non-automaton-based implementation.