Lelelo1
commented
3 years ago This will not be done with a search engine anymore, but be done with existing machine learning models, see #12.
Closed Lelelo1 closed 3 years ago
Lelelo1
commented
3 years ago This will not be done with a search engine anymore, but be done with existing machine learning models, see #12.
Identifying tags from search phrase I think the starting objective is to identify Overpass tags to understand what the searcher is looking for. It should be quick and performant in contrary of fetching a whole area of data with an overpass query, that I where thinking first, and analyse with a text engine. There is a language pre step I don't want to bring up in this issue. There needs to be well all-around working tag base is a size so that it can be embedded in the client (Or dynamically put in app storage in a initial remote fetch).
I think there should be more than one file or table. For instance countries don't seem to be part of the taginfo-wiki.db that otherwise contains good all around commonly content, being 15.4 MB. Performance-wise it's better to spawn several queries at once, so again having more files or more tables should make it faster to find the right tags for the search word made.
So, there needs to be more investigative and evaluative work made. So it there is less duplicate tags and good range of categories while still being small fast and performant.