Site search function

tnt wrote 8 years ago: 1

I'm not sure, is it a bug or a feature, but the search function is kind of weird. After a year and a half on the TVmaze, I still cannot comprehend, how it works, and why is results sometimes weird and unpredictable.

Let's say, I want to find all Marvel Comic series. I'm putting "Marvel's" in the search field, and the fun begins. First, the order of results is beyond my comprehension, looks totally random. It's not alphabetical, not by popularity (number of followers), not by year, not by channel.

For the reason above, positions 3 and 4 in the results are "Modern Marvels" (50 followers) and "Young Marvels" (0 followers), way above "Marvel's Agents of S.H.I.E.L.D." of Jessica Jones, or Luke Cage etc.

For the same reason, "Marvel's Cloak and Dagger" is not in the results at all. Because there's only 10 positions for the shows in the search results, and two of them are occupied by the less relevant entries.

The next problem, partial match. Recently I've approved a duplicate. There's the series in the development, about Sheldon Cooper, under two different names, "Sheldon" and "Young Sheldon". So, the common sense and experience with other search engines would tell you, that searching for the "Young Sheldon" will return "Sheldon" as one of the results. But not. Why? Don't know. On the other hand, searching for "Sheldon" will give you a bunch of German/Scandinavian series with unspeakable names and a word "Helden" in the name. So (Young Sheldon != Sheldon), but (Sheldon = Helden)? Weird.

Are there any tricks, any keywords or wildcards supported? Any way to get some predictable and relevant results? E.g get more relevant entries first, limit results to Show or Person, use asterisk for partial matches etc...

Thank you.

david wrote 8 years ago: 1

Search results are sorted by a mix of relevance and popularity.

Popularity isn't just the amount of followers, it also considers pageviews and favors recent activity over historical activity.

Relevance means that exact matches are always shown on top, and furthermore that a result ranks higher when there's less difference between your query and the show name. For example, "modern marvels" is a more relevant match to "marvels" (around 50% matches) than "marvels agents of shield" (only around 25% matches).

All terms in your query are used to match. Otherwise, searching for "game of thrones" would also consider the hundreds of shows with "game" or "of" in the name. That's pretty much how every search engine works, e.g. http://www.imdb.com/find?ref_=nv_sr_fn&q=game+of+random+thrones&s=all. If a show has multiple names, they should be added as AKA.

I'll definitely use your feedback for improvements in the future, but it's a much more complicated topic than you think. :) I don't think much about the algorithm should be changed, but offering you to sort the search results and scroll through them would be a better option.

tnt wrote 8 years ago: 1

Ok, thank you, it's a bit clearer now. I understand the complications and the fact, that without a more complex search language it's simply impossible to get exactly what you looking for. Let's hope that someday it'll change :)