Search function improvements

Tonks wrote 10 years ago: 1

Hi, this is what appears when you search for Chicago Fire

http://postimg.org/image/qdfm40px3/

Starring : David Eigenberg, Yuri Sardarov, Joe Minoso

I'm not sure how those actors were selected, but yuri and minoso were guest stars for the first season. And the main three actors would be Taylor Kinney, Jesse Spencer and Monica Raymund, if it were arbitrary cause that looks arbitrary to me.

Could you improve it ? I'm not even sure adding stars is needed if you have an image to separate shows with same name... I think it might be superfluous when you want results after a search but i don't mind it, i just have a question on how these actors are selected ? Number of appearances ? If so, the others have same "weight" they were in the same amount of episodes, but for instance Yuri, Joe and stolte all recurred through s1 and were added only at season 2 as main cast. So not same weight per se.


JuanArango wrote 10 years ago: 1

Tonks wrote:
Hi, this is what appears when you search for Chicago Fire
http://postimg.org/image/qdfm40px3/
Starring : David Eigenberg, Yuri Sardarov, Joe Minoso
I'm not sure how those actors were selected, but yuri and minoso were guest stars for the first season. And the main three actors would be Taylor Kinney, Jesse Spencer and Monica Raymund, if it were arbitrary cause that looks arbitrary to me.
Could you improve it ? I'm not even sure adding stars is needed if you have an image to separate shows with same name... I think it might be superfluous when you want results after a search but i don't mind it, i just have a question on how these actors are selected ? Number of appearances ? If so, the others have same "weight" they were in the same amount of episodes, but for instance Yuri, Joe and stolte all recurred through s1 and were added only at season 2 as main cast. So not same weight per se.

Hmm, Yuri and Minoso are main cast of the show , I think that is why they are showing.

cheers
Juan

Tonks wrote 10 years ago: 1

Yeah, i understand that. But why these three ? Like i said if you would follow who is more important as in the show, Kinney, Raymund & Spencer would be. Not that i'd be happy with that result by itself, but it would make more sense than 1 cast from the beginning and two that were guests all through season one before being promoted to Main Cast.


JuanArango wrote 10 years ago: 1

Tonks wrote:
Yeah, i understand that. But why these three ? Like i said if you would follow who is more important as in the show, Kinney, Raymund & Spencer would be. Not that i'd be happy with that result by itself, but it would make more sense than 1 cast from the beginning and two that were guests all through season one before being promoted to Main Cast.

I understand what you are saying, this has to do with the sorting of the main cast itself, which david still needs to tweak abit :)

cheers
Juan


david wrote 10 years ago: 1

The cast excerpt in the search results is sorted like it is everywhere else: based on the total number of appearances of a character within a show, regardless of whether the appearance was as guest or as main cast. If you click on the characters page, http://www.tvmaze.com/shows/59/chicago-fire/charac..., you'll see the details.

(In this case the top 8 characters have a pretty much equal amount of appearances, so the top 3 cast list you'll see on the search page can vary from day to day as new episodes are added.)

Tonks wrote 10 years ago: 1

Okay, i see.

I'm guessing that when you straightened the bug between main cast and guest cast, this will maybe sort itself out ? I'm looking at the page and seeing the same number of appearances for main cast and recurring is really weird.

I did the search a few times and got the same results, i wouldn't have asked otherwise because i would have considered it a fluke the first time.

But my question would be : is it even advisable to have something that varies like that ? I mean, i don't even get why a cast moves around whether on show main page/episode page etc. It's probably draining also a lot from the DB whatever system you use, if something is not static, you have request upon request every time someone loads a page and it gives a different result. Multiply this by the number of users which will only increase, that's a huge load on your servers. Traffic that will only increase, no ?

It also gives the impression that those are the stars of the show, while most people would agree they're supporting. I'm saying this looking at the result of someone doing the search and seeing these three. It's like when you search for a movie, if you give "weight" like that, maybe the bartender would be showing instead of let's say Robert Downey Jr, Chris hemworth and Chris Evans, see what i mean ? That's what the results feel like to me.


david wrote 10 years ago: 1

Let's keep this on topic: the search function.

Tonks wrote 10 years ago: 1

david wrote:
Let's keep this on topic: the search function.

isn't it part of the function to show all that info or am i missing something ? Or do you want a more specific topic about what should display for search results?


JuanArango wrote 10 years ago: 1

Tonks wrote:

isn't it part of the function to show all that info or am i missing something ? Or do you want a more specific topic about what should display for search results?

How the main cast is listed has nothing to do with the search function :) This is a compeltely different matter.
A matter that needs to be talked about tho. But we should do this in a seperate thread :)

What jaguar said that 20/20 should come up first when you search for it and not as third, this needs to be improved and this is the topic here.

cheers
Juan


MichaelDeBoey wrote 10 years ago: 1

srob650 wrote:
I don't think this can be considered an error either, there is no show called "Mad Man".

Oops my mistake --'


Trystian wrote 10 years ago: 1

Greetings all,

First post. Yay. I want to personally thank you David, and the rest of your dev team for creating such a robust, clean, and friendly site for us TV addicts, as well as a great API for those of us who also like to "dabble" in programming.

Now back on-topic...I was originally looking to fix a bug in an application I'm working on, but found this particular issue to be on your end.

-----

Issue: When searching for "The Flash" (without quotes), the first result(s) should be the latest and/or most popular by votes/traffic.

Expected results:

1. Using your main web search, I was expecting the first result to be the the exact matches in order of descending release dates, then fuzzy search results thereafter. Example: The Flash (2014), The Flash (1990), The Flash (1967), Flash Gordon (2007). URL: http://www.tvmaze.com/search?q=The+Flash

2. Using your API's singlesearch endpoint, I was expecting only The Flash (2014) to be returned. Example url: http://api.tvmaze.com/singlesearch/shows?q=the%20f...

Actual results:

1.Using your main web search, The Flash (1967) was the first result, with the rest of the results as expected.

2. Using your API's singlesearch endpoint, The Flash (1967) was the only result returned.

-----

This should be filed as two separate bug reports, but I think it stems from the same results sorting issue.

If my assumptions about the popularity of the 2014 version over the 1967 version are incorrect, or if this is by design, then please ignore this "bug" report.

Thank you,

-Trystian


david wrote 10 years ago: 1

Welcome Trystian and thanks for the kind words!

It is our intention that the most popular show comes out on top, in case there are multiple shows with the same name. This doesn't always work correctly yet though, it's a work in progress.

In general though we recommend to use the "search" endpoint instead of "singlesearch" whereever possible, so you can sort the various results whichever way you prefer. Check this thread for some background and references: http://www.tvmaze.com/threads/314/shows-using-same...


Trystian wrote 10 years ago: 1

Hello David,

You are very welcome. I also want to thank you for designing the API to return JSON instead of non-standard data formats. JSON is so easy to work with.

Actually, that was the thread I started on, which led me back here.

I realize that using the "search" endpoint instead of the "singlesearch" would be more robust and occasionally more accurate with local filtering. I was hoping to be able to retrieve the requested information in a single call. Out of about 250 unique show look-ups so far with my application in the last three weeks, this is the only one that has returned an incorrect result. Pretty impressive on your end. Now I need to eliminate about 30 unrelated bugs on my end.

Adnaan wrote 10 years ago: 1

david wrote:
Welcome Trystian and thanks for the kind words!
It is our intention that the most popular show comes out on top, in case there are multiple shows with the same name. This doesn't always work correctly yet though, it's a work in progress.
In general though we recommend to use the "search" endpoint instead of "singlesearch" whereever possible, so you can sort the various results whichever way you prefer. Check this thread for some background and references: http://www.tvmaze.com/threads/314/shows-using-same...

Hello,

I cant believe I never said this but I am extremely impressed by this site and your database format! Great work by the TVMaze team!

I was wondering, how much Levenshtein distance do you use for your fuzzy search code?


david wrote 10 years ago: 1

:-) Thanks

Automatic fuzziness: 1 for words with more than 1 character; 2 for words with more than 5 characters.

Try 30 days of free premium.