Today somebody requested MTV series "The City". The search returns nothing, there's no such series in the results. However, manually browsing all MTV series I was surprised to find, that the series actually exist.
Shouldn't the exact matches go first in the search results?
Considering that TVDb ID provided by requestor was different from the one we have, this situation potentially was leading to creating duplicate series.
Yes, exact match would have a higher score, however, the word "The" is ignored, so what you're actually searching for is "City", which is no longer an exact match.
gazza911 wrote:
Yes, exact match would have a higher score, however, the word "The" is ignored, so what you're actually searching for is "City", which is no longer an exact match.
It looks like "The" ignored only in the search query, not in the names. So it compares "City" to "The City", which gives about 50% correspondence.
Kinda awkward. What's the purpose of having search engine, that couldn't give you what you want :)
I think it would be really helpful to have some kind of "More results" button. Ideally "More shows" and "More people", to restrict the limited list of results to show only what you looking for.
tnt wrote:
I think it would be really helpful to have some kind of "More results" button. Ideally "More shows" and "More people", to restrict the limited list of results to show only what you looking for.
Yup, multiple people have said it, but so far it hasn't been implemented.
In regards to the original problem, I don't know exactly how the score boosting for exact match is done, so maybe David will be able to tell us if it would be possible for "The" not to be ignored for exact matches.
Looks like David silently improved the search :) Now it have multiple result pages, separate for shows and people. Awesome!
tnt wrote:
Looks like David silently improved the search :) Now it have multiple result pages, separate for shows and people. Awesome!
Yup, though at this point it looks like that’s just in the main site and not the API (I don’t know whether he intends to add it or not).
The "the" isn't the problem, it doesn't invalidate an exact match. The issue is more complicated, it's a result of Elasticsearch internals that aren't easy to tweak. (Specifically: the impact of field-length norms on the result's score).
So yeah, I added the pagination to the search results to help you out when a query's affected by this. :)
(I did improve the situation compared to when you reported this; the exact matches are now at least listed on the first page of results)
david wrote:
The "the" isn't the problem, it doesn't invalidate an exact match. The issue is more complicated, it's a result of Elasticsearch internals that aren't easy to tweak. (Specifically: the impact of field-length norms on the result's score).So yeah, I added the pagination to the search results to help you out when a query's affected by this. :)
(I did improve the situation compared to when you reported this; the exact matches are now at least listed on the first page of results)
This is very helpful, thank you!
