F. Version History

First public release (6 September 2002)
Modules for data cleaning and standardisation, hidden Markov model (HMM) training.
Second public release (14 April 2003)
Complete object-oriented re-design, includes modules for probabilistic linkage.
Updated second public release (26 June 2003)
This version is published under an updated ANUOS license version 1.1. Added top level routine standardise in module febrl, added the flexible classifier in module linkage, added a PassFieldStandardiser which allows simple passing of fields without standardisation, added the data set generator generate.py, added improved one-to-one assignment based on Auction [3] algorithm, improved memory management and verbose output, added process indicators, added Berkeley database library support in Shelve data set, plus many smaller improvements and bug fixes.
Bug-fix release (November 2003)
This is a bug-fix release for Febrl 0.2.1. Thanks to everybody who sent us bug reports. The manual is now processed using the Python 2.3.2 documentation system.
Third public release (April 2005)
Includes a geocoding matching system based on the Australian G-NAF (Geocoded National Address File) database. Various smaller additions, including several new field comparison functions, a telephone number standardiser, CSV (comma separated values) output of the linkage results, and a new logging system based on the standard Python logging module. Updated all modules to Python 2.4.