0
How can data from one article be used for cluster size selection and parameter tuning for another article if the rows of the feature matrices (n-gram frequency matrices) across articles do not correspond with each other?

In the selection of number of clusters and in parameter tuning, how can data from one article, say Article 3, be used for cluster size selection and parameter tuning for Article 6 if the rows of the feature matrices (n-gram frequency matrices) across articles do not correspond with each other? (For example, the first row for Article 3 data corresponds to abducted while for Article 6 it is abolition. The sizes match but the actual features don't.)

waiting for moderation
1 Answer
0
Accepted answer

Our approach can be considered as a weak parameter tuning though cause we tune our models in different datasets but for closely related tasks. We followed this protocol because our data are relatively small.

waiting for moderation
Other questions about this article