[Corpora-List] unsupervised with semi-supervised

[Taras Zagibalov]

Hello and many thanks to all you who replied to my question regarding 
unsupervised and semi-supervised learning and provided me with 
interesting and valuable information.
Still I found some definitions quite ambiguous and I would like to 
clarify a few things.
I'd appreciate if you could share your ideas about the following examples:
1. System A uses one million labelled examples for training. It works 
iteratively and can also use information obtained for self-training. 
System B uses only 100 labelled examples and also can use information 
obtained from unlabelled data for improving performance. Does it mean 
that both of the systems are unsupervised?
2. System A uses a seed vocabulary consisting of 100000 items for some 
task. System B uses only two seeds. Are they both unsupervised?
3. System A uses a big set of rules (hundreds) describing a language. 
System B uses only two rules. Are they both of the same kind (supervised)?

More generally speaking, does the amount of manual input matter for 
deciding if a system is (un- /semi-) supervised?

Best regards,
Taras Zagibalov
University of Sussex



_______________________________________________
Corpora mailing list
Corpora at uib.no
http://mailman.uib.no/listinfo/corpora