BibAuthorID: improvements, fixes and optimizations
- Now it's possible to disambiguate surname clusters without precomputing clusters for every surname, makes testing and tweaking much faster
- Disable citation analysis for INSPIRE disambiguation, too big memory footprint for such a small increase in accuracy
- Tweaking: facilities to collect and analyze statistics about final results in disambiguation. Allow easy plotting of results features against thresholds and so on
- Added lots of assertions and tests to allow early detection of insidious problems during disambiguation
- Names comparison now translates unicode to ascii, thus ignoring accents. This improves disambiguation and search.
- Outdated claims are now correctly handled by the interface, and present message boxes to the user in order to make clear what is happening
- Greatly improved debug prints to ease debugging
- Comparison functions now behave properly and offer always correct results in correct range
- Comparison function fixed bugs in handling of INSPIREid
- Redistribution of weight in comparison function to achieve more significative results
- Disambiguation computations redefined in order to limit operation overflow/underflow problems
- Paper claiming code now handles safely all known stressful situations dealing with outdated data
- Admin paper claiming interface offers the possibility to manage external ids for persons. Kind and features of external IDs are configurable through config file.
- UI style fixes
- UI fixed various bugs (mostly javascript related)
- UI improved extendability and reusability in templates through verbiage dictionaries
- UI Disabled option to 'forget decision' on rejected papers which made no sense
- Added a prototype for export of personID information in XML format. Usable but unfinished, will probably be moved to bibexport.
Co-authored-by: Nedko Nedkov <nedko.stefanov.nedkov@cern.ch>