On 15.11.2013, at 02:45, Swirl <[email protected]> wrote: > Richard Eckart de Castilho <rec@...> writes: > > >> In ClearTK, in particular in the machine learning module, there were >> analysis engines for learning classifiers. However, there wasn't one >> AE per training algorithm, rather one passed in the name of a class which >> implemented such an algorithm was passed to the AE, along with any > parameters >> that the algorithm would require. The AE would then instantiate that > algorithm >> and pass parameters on to it. In order to tell apart which parameters were >> meant for the AE and which were actually meant for the instantiated > algorithm, >> the prefix was used. > > Thanks for the detailed explanation. > > BTW, out of curiosity, does that mean that your algorithm are implemented as > shared resources which are passed to the AEs?
ClearTK appears to have recently switched to uimaFIT 2.0.0, but I didn't check how they handle this now. In DKPro Core and DKPro Text Classification, use shared resources to implement a strategy pattern for components, yes. -- Richard
