It is like Daniel says and it is good to have training data that is close to the data you intend to process with the model.
Jörn On Fri, Oct 6, 2017 at 5:32 PM, Dan Russ <danrus...@gmail.com> wrote: > I believe it does. Every word is classified as “begin”, “inside”, or > “outside” - BIO encoding, so an event is generated for “she” and then “does” > and then “not” — all of which is classified as “outside”. > > Anyone smarter have a comment on this??? > Daniel > > >> On Oct 6, 2017, at 11:26 AM, Benedict Holland <benedict.m.holl...@gmail.com> >> wrote: >> >> Hello all, >> >> I am working on getting together a file with a list of tokenized sentences. >> I have a quick question: >> >> Can name training data contain sentences without any tags? >> >> For example, if I had a sentence like >> >> <START:person> Molly <END> enjoys pancakes in the morning . >> She does not enjoy being woken up at 4:30 by her cat . >> >> Does the second sentence provide any additional benefit to the ME model? >> The answer to this question should probably be in the documentation. >> >> Thanks, >> ~Ben >
