Yes, you can either train a model that recognizes both types at once or
filter out the tags you don't want to be trained on.


On Thu, Jun 12, 2014 at 3:34 AM, Saurav Sharma <[email protected]> wrote:

> Hi all,
>
> I am new to this realm (so I apologize for the n00bness) and had a question
> regarding building the training set for NER Models.
>
> Let's say I want two models, one for finding Brands and one for finding
> Weights.
>
> Would it be valid to have training data full of data that is double tagged?
>
> For example:
> <START:brand> Pop-Tarts <END> Brown Sugar 12 pk <START:weight> 21 oz <END>
>
> I want to be able to identify the brands and weights. Can I use a set
> filled with the above to train a model specifically for "brand", and then
> use the same data to train for "weight"?
>
> Thanks in advance!
>

Reply via email to