We've sent a verification link by email
Didn't receive the email? Check your Spam folder, it may have been caught by a filter. If you still don't see it, you can resend the verification email.
Started April 17th, 2006 · 41 replies · Latest reply by Bram 18 years ago
is there a complete list of TAGS?
I've seen this...
http://freesound.iua.upf.edu/tagsView.php
is it all inclusive?
I was wondering if there's a way to organize the TAGS in a
Hierarchical tree model?
Is that discussed much?
Kelargo
right now there's no way to view all the tags, but you can search for a tag though.
a hierarchical tree, well, if you feel like trying i can send you the full list of all tags by email + sound they belong to.
but I think the whole point of tags is that they AREN'T hierarchical!
- bram
I too have bent Bram's ear on the theme of tags. To me it's a shame that (for example) a search of tags and descriptions for "speak" returns a different set of results to a search for "speaking".
Some kind of root word dynamic lookup would help greatly.
Also, I can't see at the moment why a search for "speak" returns "lg water4.wav" and "lg wind2.wav"
Mike
the whole point of tags is that they are free and up to the user. instead of enforcing behaviour (with predefined lists of tags, or trying to group tags before they are used), behaviour is emergent. this is the very foundation of folksonomy ( http://en.wikipedia.org/wiki/Folksonomy )
There is a sort of root word dynamic lookup in the search right now, and it's exactly what is getting you the false result right now. We use Wordnet ( http://wordnet.princeton.edu/ ) to link certain words together, but this is sometimes very bad. An example. Wordnet defines "honky" ( http://vancouver-webpages.com/cgi-bin/wordnet.html.cgi?Honky ) as "whitey" or a "offensive name for White person". This is alright in it's context, but very bad for freesound. "honky tonk piano" will result in a subsearch for "white".
Another example where you will see context is everything. "Piano." The instrument, or the method of playing (pianissimo)?
When you search for speak, you get a subsearch for "murmur", check LG's descriptions.
"Intelligent" searching (i.e. semantic searching) is a very difficult subject. And, look how google solved it: they didn't. Searching for sax of saxophone will NOT give you the same result on google.
- bram
Like in playing that children's game... "is it animal, mineral, or vegetable"?
then in twenty guesses one tries to figure out what it is...
I dont think a strick all-inclusive hierarchical tree is practical...
and the looseness of tag-ing has much value, as you described...
But I would love to see the list of tags!
could you please send it to me?
Thanks for posting the list of TAGS.
I would not have guessed some of the TAGs...
It helps me explore the sounds more.
One more question:
Is the list dynamic? I know its PHP generated...
but I was wondering if the list itself is dynamically generated...
Or statically cached on a daily bases or such?
Also, thanks for posting the reference on Folksonomy.
I never heard of that before... Helps me understand the logic.
Regards,
Kelargo
Cool! This will help in finding/sorting through relevant tags.
I think folksonomy can be a double edged sword... the advantage of taxonomy is that all things that should be in the same category are, at least in the mind of whoever the authority figure is. With folksonomy, you can get sounds that should have the same tag, but don't... and it takes extra work to try and make sure you're grouping your sound with other sounds that are relevant. So, the advantage is in unique and adaptable categories, but the disadvantage can be in too many categories (especially variations in spelling, spacing, or pluralization.) Ex. "chord" and "chords", "click" and "clicks", "fx", "effect", and "effects". To eliminate this, I think that there should at least be some general guidelines for writing tags, although i've never seen any sites that have guidelines. del.ici.ous does have tag suggestions though, which is another good solution.
BTW Bram, google does have this functionality if you use the ~ operator. For instance, result #3 for "~sax" is a page on saxophones, whereas result #3 for "~saxophone" is a page on the SAX API. I agree that it's not perfect though... for instance, nobody searching for "~saxophone" will want to read about the SAX API.
Now we're getting into interesting discussion land...
Halleck
I think folksonomy can be a double edged sword... the advantage of taxonomy is that all things that should be in the same category are, at least in the mind of whoever the authority figure is. With folksonomy, you can get sounds that should have the same tag, but don't... and it takes extra work to try and make sure you're grouping your sound with other sounds that are relevant.
This is why Freesound let's you tag sounds that aren't yours. You can add tags to any sound! No-one (*no-one*) uses this feature, but it couldbe interesting forming a team of "pro" taggers in Freesound that establish a good way of going through the samples and tag them. I could even create an easy interface for it. But, .... this takes time, so it should be organised in some way.
Halleck
So, the advantage is in unique and adaptable categories, but the disadvantage can be in too many categories (especially variations in spelling, spacing, or pluralization.) Ex. "chord" and "chords", "click" and "clicks", "fx", "effect", and "effects".
I personaly do NOT think click and clicks is the same. Neither is chord and chords! One implies 'one' instance, the other many instances. And looking at even only first page of sounds tagged with 'click' and 'clicks' it immediately shows that people do see this distinction when tagging:
clicks : http://freesound.iua.upf.edu/tagsViewSingle.php?id=9
click : http://freesound.iua.upf.edu/tagsViewSingle.php?id=64
That said, it might be interesting to SEARCH for pluralizations when searching. But, this needs a robust pluralization algorithm or dictionary, something which isn't easy to find.
Halleck
del.ici.ous does have tag suggestions though, which is another good solution.
I'm sorry but this isn't true. del.icio.us only has tag suggestions when you are tagging an article which has already been tagged! I.e. 500 people add Freesound and tag it as "sound", if you come to del.icio.us and add Freesound, it will suggest you to use "sound" as a tag. This is very different. There is a way to suggest tags via similarity though: a user uploads a sound, similar sounds are found in the database and from this a set of "similar" tags is selected.
But this method is very far from perfect, as you can see here:
http://freesound.iua.upf.edu/autotag.php
Halleck
BTW Bram, google does have this functionality if you use the ~ operator. For instance, result #3 for "~sax" is a page on saxophones, whereas result #3 for "~saxophone" is a page on the SAX API. I agree that it's not perfect though... for instance, nobody searching for "~saxophone" will want to read about the SAX API.
True, but only very very few people use this functionality (I admit I do from time to time)!
- bram
BramThis is why Freesound let's you tag sounds that aren't yours. You can add tags to any sound! No-one (*no-one*) uses this feature, but it couldbe interesting forming a team of "pro" taggers in Freesound that establish a good way of going through the samples and tag them. I could even create an easy interface for it. But, .... this takes time, so it should be organised in some way.
- bram
I think we had a discussion last year that could relate to this topic: the matter of English being the official language here at FS (sorry that I always come to this issue, but for some of us writing in English represents no minor effort, I for one often need a dictionary to find the word, albeit approximate).
Language is, IMO, allimportant when it comes to archiving, including into databases, etc. Bram, I mean these "pro" taggers you suggested should necessarily be English-speaking (since they only possess the vast repertoire of words you need to describe sound). For example, when acclivity says something sounds "squidgy" I feel a bit stupid because all I know is that there is an English word for "squid" (calamar in Spanish), but "squidgy"??? After some effort, I now associate the Spanish words "churretoso" (greasy) and "resbaloso" (slippery) to squidgy, but I will never be sure about my personal translation.
In short, sound (and music) actually break the border of languages. But when we associate words to sounds the borders are rebuilt, so ... don't expect good tagging from many of us...
I'd be happy to help with tagging. Have good command of English (apart from "squidgy"!)
but not so good on technical sound recording terms.
Browsing the complete list of tags (just under 4300 of them) my initial reaction is
that all hyphenated tags should be split into their separate components.
e.g. who is going to search for "indian-music-instrument" but one may well search
for "indian" or "instrument". Similarly "australian-animals", "black-metal" ...
I'd be happy to sort out the obvious spelling "miss steaks" - would you like me
to start that process, or do you want to control it in some way?
e.g. "announcment", "ansering" etc.
Mike
p.s. Squidgy is indeed not in the dictionary. Not even squigy. Something that feels
squidgy would be like a jellyfish maybe. Squidgy could describe the noises your
tummy makes when digesting food ... or those 'orrible artifacts in low bitrate mp3
compressions. (like I once upon a time USED to upload!)
See "Jellyfish.wav", "Trans02.wav" - these are squidgy sounds.
Mike
p.p.s I see little difference in a tag search for "click" or "clicks" - I think a distinction
here is splitting hairs - and assumes the words are nouns - what if the user meant
the verbs? Then, I submit, there's no difference.
Mike
acclivity
Browsing the complete list of tags (just under 4300 of them) my initial reaction is
that all hyphenated tags should be split into their separate components.
e.g. who is going to search for "indian-music-instrument" but one may well search
for "indian" or "instrument". Similarly "australian-animals", "black-metal" ...
All, no. Some, perhaps. Field-recording is quite an interesting tag.
acclivity
I'd be happy to sort out the obvious spelling "miss steaks" - would you like me
to start that process, or do you want to control it in some way?
e.g. "announcment", "ansering" etc.
Spelling mistakes, I certainly agree they should be removed.
acclivity
I see little difference in a tag search for "click" or "clicks" - I think a distinction
here is splitting hairs - and assumes the words are nouns - what if the user meant
the verbs? Then, I submit, there's no difference.
Look at the wave forms. "Click" is most of the time a single click, "Clicks" are almost always more than one click.
- bram
I am finding this list to be very interesting and helpful. I am wondering if it would be possible to generate an alphabetical list (currently it is generated by number of sounds per tag, right?) I think alphabetical might be useful for finding similar terms, "click" and "clicks" and "clicking" or "analog" and "analogue" for examples. And I would be happy to donate some time for tagging/spelling correction.
Tim
( Mike,
I have another great example of automatic (de)pluralization which gets you into trouble.
shutter : http://freesound.iua.upf.edu/tagsViewSingle.php?id=1480
shutters : http://freesound.iua.upf.edu/tagsViewSingle.php?id=4144
)
I added a parallel list alphabetically ordered. No idea what those few things at the start are Those are definitely errors!
By the way, if you guys start working on this list, please make sure you always note down the ID of the tag as well as the tag and the reason why it should be joined/corrected/.....
The thing I see that could be done are:
1. joining two tags that have exactly the same semantical meaning (this is a difficult one)
2. joining two tags, one misspelled (how to prevent more misspellings? )
3. removing tags (weird and useles)
- bram
Wow, that was really fast. Thanks.
As far as editing/changing tags, do we send those to you to be approved? Perhaps I'm just dumb, but I can't seem to figure out a way to edit them. (I was just going to fix a couple I saw misspelled, like "ansering-machine" or "rhtyhmic". Sorry to be annoying, but I just finished a big job so I have free time (and sounds to post).
Tim
(Bram: I'm sure there are more exceptions that prove the rule ...
but that's it - they are exceptions.
Don't throw the bathwater out with the baby, or some such inverted saying.
Mike)