Sep 4, 2018 · 1 min read
Hi Giovanni. Have you tried combining both, user-inputs and document-data? I did not try it yet but I assume a neural network having access to
1) search-inputs that lead the user to a certain document
2) document-data (titles, descriptions etc)
could combine both metrics to produce good search tokens. At the very least 1) could be used to set the ElasticSearch-weights for an index made from 2). What is your opinion on this approach?
