How to trace back to the original Document in "009007_SentimentClassification" example

   We noticed the 1st row0 in "Decision Tree Preditor" node is mis-classfied(it should be NEG, but POS is predicted), so we want to trace back to the original document, could you show us how to find the row in IMDb-sample.csv?

Thanks in advance.

Hi,

you can set e.g. the row id of the original data set (IMDB-sample.csv) as source of the document. This meta information will be carried along in the document. After creating the document vectors there is still a document column. With the Document Data Extractor you can extract the source field (which is the row id) and join back the original data set.

Cheers, Kilian