search does not return partial text match results

sdebodt commented 2 years ago

Describe the improvement When searching the publications using the search box, I do not get results when the search term as part of a word of e.g. the publication title. For instance, when I search with "della", I do not get the publication that has the word "DELLAs" in the title (Biblio ID: 4419918). (The result I get has the complete word "DELLA" in the abstract, so that is correct)

To Reproduce

Go to 'Publications overview' page
Enter search term in search box
Click enter
Check results

Expected behavior I expect to be able to search with part of words as well as complete words.

Screenshots part of word results:

complete word results:

sdebodt commented 2 years ago

Dependency on LibreCat database. To be revisited later.

nicolasfranck commented 2 years ago

Note to self: use ES internal field _score to sort on

mietcls commented 1 year ago

@nicolasfranck is this something we can do now?

nicolasfranck commented 1 year ago

@mietcls it's doable, but it requires

adding an extra edge ngram field, that contains data from other fields. @nics I see that fields contributor.ngram and contributor.phrase_ngram contain contributor names, and are now used during search. Why not a common field for ngrams (and that contains names and titles)? We already have a common field all that now contains titles and contributor names only. It would mean though that we cannot distinguish between "match on title" and "match on contributor" when it comes to sorting. But I guess scoring should be very low for ngrams anyway?
probably more data, and indexing time. Anyway scoring should be very low, and minimum length not too low (i.e only when words are longer than 2 characters), as it would bring "crap" results on top. Also creating ngrams from several sources should be created with caution. We already search in multiple fields, so that someone may enter both an author's name and part of the title (e.g. Plato politeia). I've tried this in the past, and saw that people where confused on why they got results. The more fields you add, the greater the confusion becomes

mietcls commented 8 months ago

Closing this since it fits into a bigger search question.

ugent-library / biblio-backoffice

search does not return partial text match results #89