Closed p1d1d1 closed 1 year ago
The issue described might be related to stemming. @eliaferrari is that something you could check? Maybe we have to scale back stemming :/
We could use stemming on the query, but I think the issue is more complex than just that. The extended search will require additional database columns in the database (I've already implemented some functions in #11 ) as well as a ranking method/function, which can be derived from different search methods.
We found out that the issue is not due to stemming but due to redis checking for an exact match. Instead we would need a LIKE or CONTAINS operation, if that is possible with redis. Needs investigation
The search function in redis is now including as many results as possible, which are subsequently postprocessed with a ranking function, sorted and finally returned to the API with additional ranking information.
We need to improve the search. Currently, e.g., if you search for "swissboundaries" you only get 2 entries and datasets containing "swissboundaries3d" are not returned. Another example: if you search for "grenzen" you don't get the datasets from Bund (like Landesgrenzen, Gemeindegranzen, etc.).
We need to be able to provide the same results (if not better) as https://davidoesch.github.io/geoservice_harvester_poc/. I guess we can achieve this if the search go also through the abstract and is additionally able to recognize parts in composed words.