Freesound Forums

Seeing title "Celebraing Diversity" in one forum posts, I realized what bugs me about freesound. Search results come sorted by nr of downloads by default. Of course you can change it, but how many users do? Not many. That means sounds that had been downloaded the most are downloaded more and more.

It is also easy to forget the search is sorted this way and even you don't want to use the most used, you've allready downloaded very popular sound .

The random sound of the day won't help to balance things out. New sounds on the front page do some work but in the end it's just a small drop in the sea.

So what do you think? Is sorting by nr of downloads by default a good thing?

RANDOM is fair to EVERY user.

I'm not sure that the database counts correctly.
For example, doing a search for "fluid":
First result 2240, second result 2569.
Searching for "even":
First result 60651, second result 18750, FIFTH result 23522.
Am I missing something?

But more on topic, I agree that there could be a more fair way to return search results. On the other hand people are quite lazy and want someone else to pick what is best/popular for them.

1) it's nice when you don't hear the exact same sounds everywhere / like some overused door samples in the TV shows

2) sounds with many downloads are also often one of the oldest on the server and not allways the best quality (also because some more pro soundrecorders joined freesound later...)

I have to say you've surprised me Corsica_S with the "Fluid" search result. I've never noticed the results are not 100% sorted by the nr of downloads. There is something that mixes them a little. But it is 90% dependent just on nr of downloads.

Search results for "door" give you 3331 sounds. The first item is form 2006 http://www.freesound.org/people/FreqMan/sounds/23168/ and has been downloaded over 26000 times. It has been recorded with SB Audigy and mackie mixer (and sounds like it). I have nothing against that sound, it can probably work at many places, but other results drop in dl nr very fast. I have this impression about big part of the search results I saw. Everyone downloads the first one, and maybe some other down the line and of course no one goes to page 223 and even less people to page 140 .

Other problem that I can see right away. Result #4 on the first page is lightning_strike.wav, why on earth? 15724 downloads and there is "Lightning strikes the house next DOOR, and echoes..." in the description. So here something that's not even the sound you're searching for and it was brought very high in the results. I know where it came from, but then you get also a lot of doorbells which aren't recognized as less relevant search result...

Maybe this issue is more about how the search engine works. It works fine when there are few results (two-three pages...), but with bigger nrs lot of the great content is just hidden away and it is hard to get to.

There are also great features of the search page, the whole right column (with filters for sample rate, licence...) is great.

Don't get me wrong, this is no moan. I just wanted to start discussion to see what other freesounders think about it.

PS. klankbeeld - yes, random (sound of the day) is fair, but it's too small "instrument" to shuffle the cards. Maybe if there were ~3 random results (in separate box) at the first page of the results followed by "most downloaded" thing it could help to diversify the lazy user's choice a bit.

There is also the "random sound" button. I often click through random sound after random sound. I find lots of great sounds that way.

I am in general agreement with the theme of this thread. I don't do many searches, but I've noticed the same trend with "most downloads first".

I would suggest a slightly more complicated search algorithm, perhaps called "most popular" which translates to ordering the list by "latest first" then filtering to re-order the priority towards "downloads-per-day". How the bias between "latest" v "downloads-per-day" was calculated might be a matter of trial and error which needs tuning for an optimal balance, but that would only require changing a single constant value once the code for the search algorithm were written.

This could at least address the problem that the a sound that has been around for years is bound to have been downloaded more times than one which was uploaded last week, giving a fairer exposure to it's potential.

I think even a simple filter based on "Number-of-downloads" divided by "Number-of-days-in-database" would give much more usable results even if it didn't undo the bias of older sounds being preferably downloaded due to their constant appearance at the beginning of the search results.

Well..., that's what I think!

Wibby.

I searched on freesound today and to my surprise, the results were more relevant than before. So I checked the sorting and there it is "automatic by relevance". My first impression? Great! Thank you.

matucha wrote:
Seeing title "Celebraing Diversity" in one forum posts, I realized what bugs me about freesound. Search results come sorted by nr of downloads by default. Of course you can change it, but how many users do? Not many. That means sounds that had been downloaded the most are downloaded more and more.

On another point - actually, the reason for my reply to you - if you were slightly annoyed by the most downloaded sounds getting downloaded more and more.... you will probably like one of the upcoming dares...

I've never expected someone would actually listen to what I wrote/said. But here we go . Hopefully it won't get to the point where the relevance factor stays the same and just some other sounds get to the top and stay there.

Dares are nice concept, when I have spare and the theme suits me, I'll participate gladly .

happy holidays

Matous

Well, the search can include as many words as you want. It is up to the user to control that.

If you search for "gunshot" and there are 100 samples including that word as a tag or in the description, some criteria might be used to sort them further. I would assume that the sounds with "gunshot" as a tag would appear as more relevant than the ones that have "gunshot" in the description.
But for all results with "gunshot" as a tag, the algorythm might use further criteria (age of sample, number of downloads, etc to sort further).

The point is. If you search for "gunshot revolver 45" - you should get samples better sorted by relevance for what you are looking for.
Searches with more words should always produce better results.

Yep, no argument about that.