Hi Neeraj, I believe that the Document Categorizer picks the best label for document. It sounds like you want multiple classifers (if you have each sentence a document and you want to classify the sentence) OR use the NameFinderME to build a sequence classifier that can identify 1)sarasm, 2)surprise … Daniel
> On Jul 20, 2017, at 8:53 AM, Kulkarni, Neeraj <neera...@gmail.com> wrote: > > Hello all, > > I wondered if anyone had insights on classifying data through multiple data > sets or through a single data set for document categorizer ? > > I am building a sentence category classifier (e.g. surprise, disgust, > sarcasm, unknown) and wondered if one DocumentCategorizerME instance with > training data with all four types is more effective (with confidence > threshold of say 0.25 to decide on a category) or if I should seek to > categorize sentences through three DocumentCategorizerME instances (one > each for surprise, disgust, sarcasm, with confidence threshold of say 0.6 > to decide on a category, unknown otherwise). > > I am a newbie to this mailing list, and apologize if this question is > irrelevant. Please help me by pointing to the right direction. > > Best, > Neeraj
