aadl / arborelastic

Custom Elasticsearch module for Drupal 8
1 stars 0 forks source link

Search Configuration issues #16

Open elitrichous opened 6 years ago

elitrichous commented 6 years ago

Dumping some questions and concerns in here

Special characters: need to ignore commas, apostrophes, accents:

commas: https://aadl.org/search/catalog/adams,%20douglas?author=adams,%20douglas and https://aadl.org/search/catalog/douglas%20adams?author=douglas%20adams should not have a different result set. could have different relevance results? but not sure term order mattering is what patrons expect.

apostrophes: https://aadl.org/search/catalog/o'connor?author=o%27connor 166 hits https://aadl.org/search/catalog/o'connor?author=oconnor 0 hits

accents: https://aadl.org/search/catalog/caf%C3%A9?title=caf%C3%A9 vs. https://aadl.org/search/catalog/caf%C3%A9?title=cafe these probably should not get different results

All vs. Any We used to search all by default, and fail over to any if there were no hits on all. Is this still possible? It's a good solution for clueless searchers who get overwhelmed by too many hits, and also obscures weak relevancy ranking

ex. https://aadl.org/search/catalog/so%20long%20and%20thanks%20for%20all%20the%20fihs?title=so%20long%20and%20thanks%20for%20all%20the%20fihs this should work. no hits. Does elastic support did you mean?

What is the difference between the suggestions ranking / search approach and the full search? There are numerous situations where when you hit return, the thing you're looking for is the top title suggestion but does not show in the first page of search results. example, keyword search fire and fury. three suggestions are shown. the important one is #1, but the other two are #9 and #24.

Overall I think the suggestions is returning more reliable results than full search.

Index-specific searches should probably be full match by default. example: https://aadl.org/search/catalog/alexander%20mccall%20smith?author=alexander%20mccall%20smith brings up 1245 hits, but only 157 actual titles with that author. Could be another all / any switch issue.

Again order of terms and comma operator may be getting in the way. actual items by this author: 157 author: mccall smith, alexander: 645 author: alexander mccall smith: 1245

These should be relatively low-hanging fruit. Further down the road, I would like to get authors, subjects, and call numbers into the suggestions index, with the "author" "subject" or "call no" as the format for those suggestion hits, linking to an index search by that term.