The retrospective conversion is one of the most important
library automation projects of the recent years. The integrated library
systems, and their additional applications have to produce new solutions. The
circumstances have been changed radically in the past two decades. The costs
has became high, the shared cataloguing systems, the facilities of data
transfer, and the amount of bibliographic data of the electronic librarian
catalogues have became better, and greater, so the chance of successful data
retrieval and import are very good. To scan, recognise and to segment to
bibliographic data of an RDBMS the rest of paper based catalogue records seems
to be too expensive, makes a plenty of mistakes, and causes inconsistency of
the native library catalogue. It is worth to import bibliographic data from outside
instead of making a new bibliographic description. The most important data of
the library item has to be detected and to be given to a MARC record harvesting
system, to search and load bibliographic records already made by other
librarians. This detecting can be made by human, or can be made also by a
software too. If the cards are scanned, it can be done by members of a network
too. The MARC search application can recognise the same entities of the
harvested data, and give always only one record by each of them, ranking them
by adequateness too. The library staff can registrate the imported data at the
end of the workflow.