I’ve cross indexed the music4dance catalog with the EchoNest database and exposed some new features.
The most fundamental improvement is that EchoNest provides tempo and simple meter information. So I’ve been able to add tempo information to a number of songs that I didn’t have cataloged that way before. And I’ve also added the time signature tags to these songs.
But more importantly, EchoNest does some interesting analysis of the music to come up with some acoustic attributes to describe a song musically. They are:
- Beat: An attempt to characterize the strength and consistency of the beat.
- Energy: The intensity and power of the music – this is probably the most intuitive of the attributes.
- Mood: A measure of the mood from positive or happy down to sad or angry.
I am representing each of these attributes as a graphical column headed with an appropriate icon (drum for beat, flame for energy and smile for mood). Each column is sortable both from high to low and low to high and when I have data for a song, the icon in that song’s row is ‘filled’ to a proportional amount. EchoNest provides a number from 0 .0 to 1 .0 and I translate that into icons that are 0 to 100% filled (in 10 bands).
More information on this integration can be found on the help page. Or just head over to the music library and start experimenting.
For instance, here is a snapshot of a list of songs that are tagged as “First Dance,” can be danced to some form of Foxtrot and contain the word “Love“. They are sorted by “Beat” from strongest to weakest. If you’re looking for a song to dance your first dance to and aren’t an experienced dancer, you probably want a strong beat.
What do you think? Are there other acoustic attributes that I should include? Would you like to be able to sort on multiple attributes on the same time or filter on one attribute and sort on another? These are all entirely possibly, but I need your help to prioritize these features. Please feel free to reply to this post or leave feedback with any thoughts you have on this set of features.
One of the things that I’m struggling with in the music4dance project is the pull between finding lots of recommendations for songs to dance to against the desire that those recommendations being in some sense ‘good.’ As I noted in my last post, more manual curating will certainly help.
Another way to attack this problem is to tune down the mechanical methods I’m using to find recommendations. So for now I’ve changed the default for most dance recommendations to only pull up songs that someone has explicitly noted are good to dance to that particular style. So, for instance, the Carolina Shag catalog doesn’t get polluted by songs that someone has tagged generically as ‘Swing‘ that happen to be a reasonable tempo to dance Carolina Shag to. And the top 10 lists for individual dance style pages are filtered the same way.
Hopefully this will give you a better start at ‘core’ songs that work for each dance style. Check it out and let me know what you think.
P.S. What if you’re interested in getting the broader range of recommendations? Say you’re looks for songs that might work to dance a style to that might not be quite the musical style that is normally danced to. There are a couple of things you can do. First, you can go to advanced search page and check the “Include Inferred” check-box. This will give you the old results where we pull in songs that are marked with a related dance style and are of an appropriate tempo. Or you can just do a tempo filter on the advanced search page and see what comes up…
A number of people have pointed out that my attempt to generalize the “rules” that I use to pick Ballroom music, especially slightly edgy ballroom music have caused the social music parts of the music4dance catalog to veer a bit (or more than a bit) off course. One way to solve this is to spend some time on each of the social dances and see if I can get some more concentrated people knowledge to help contribute to a better list.
And since I’m going to be taking Bachata lessons for the first time starting next week, that seems like a great place to start. I’ve pulled together a couple of the albums that the teacher recommended and some Bachatas from a few other sources to get an initial list together here. What do you think? Am I anywhere close to a decent list? What am I missing or what is on this list that you absolutely wouldn’t dance Bachata to?
The other thing I noticed is that it looks like Bachata has a broad enough history that it may make sense to either split them up (like I did with Tango) or maybe more reasonably start tagging them by sub-style?
Please feel free to respond to this post with song ideas or more general suggestions. You’re also welcome to sign in to the site and start voting on songs. And if you’re not a Bachata expert, that’s all right, there is plenty of room for improvement elsewhere in the catalog.
One of the first things that I wanted to do with the music4dance project was to build a system where dancers could vote on whether a song worked for a particular dance style. That morphed into something more like “do you like to dance style x to song y.” Which still seemed like a pretty reasonable system. And I started using heart symbols to show like/indifferent/don’t like for the dance style entries on each song.
But then Amanda (the music4dance intern) pointed out that there wasn’t any way to actually say that you liked or disliked a song. So we implemented a top-level like/indifferent/don’t like for each song and I used the same heart symbols. But then things got confusing, since you could vote to like to dance a particular style to a song as well as make a song as something that you like.
I still think that there is a value in both of these attributes. For me the overriding value of the like/dislike a song is to be able to dislike a song and not have to see it again. Especially for a song that I may have particular bad associations with – a song that was massively overplayed in a dance studio, for instance. These are particularly troublesome as they are probably overplayed because they are well suited to a particular dance style, so I certainly don’t want to exclude them from other people’s searches by voting them down for matching that dance style.
In any case, this came to a head while I was implementing the quick voting mechanisms where you can like/dislike a song anytime you see it in a list – any kind of search results or dance style details pages. Once I got that feature in, I saw that it was so much easier to like/dislike a song than it was to vote on whether you find a song dance-able to a particular style. And that made me sad, since I think the dance-able vote is much more interesting (and also why you would be looking at a site like this rather than a purely music rating/referral site).
So I added in the voting link to the songs in any list where there is a single dance being listed. This includes the dance style pages as well as any searches where you choose just one dance to search on (like for instance, this list of all of our East Coast Swing songs). And while I was doing that, I started using the shoe symbol rather than the heart symbol for voting which helps with the ambiguity.
Well this all works, and I’m reasonably happy with the results. But I’m not sure I’ve minimized the number of clicks that you would use on average. Should I assume that you like a song if you vote it up as a good Cha-Cha – you could still explicitly dislike it for those cases where you actually don’t like the song even if you agree that it’s a particularly good song to dance the Cha-Cha to.
Why don’t you give it a try and let me know what you think. Go to your favorite dance page (find them here) and try both voting on the songs as dance-able to that dance (by clicking the shoe icon) and liking/disliking them as songs (by clicking the heart icon) and let me know what you think. Feedback is always welcome via the music4dance feedback form or our company email firstname.lastname@example.org.
What better than a top 100 list to end the year? Since music4dance is about the intersection of music and dance, I’ve taken the Spotify top 100 songs of 2015 (for the USA) and cross referenced it with the music4dance catalog.
About of a third of the songs were already in the music4dance catalog and had been matched to dance styles. Most of the others were songs that I could imagine partner dancing to, although some were a stretch. I made a lot of use of the “Unconventional” tag to try to show that many of these songs don’t exhibit all of the traditional attributes of the music that these dances grew up with. But that’s part of the fun of this project, and dancing in general – testing the limits of how music and dance fit together.
And of course, not all music is particularly suited to dancing of any kind. So my version of Spotify’s top 100 list ended up being only 87 songs because those were the songs that I could match any kind of partner dance, even stretching the traditional definitions.
Here’s a link to that list. You can build this link yourself (and try some variations) by going to the Advanced Search Form, clicking on the brown pencil (other) tag button in the “include tags” row of the form. Then choose tags to include – in this case I chose “2015” and “Top 100“. Clicking the search button will show the first page of the top 100 (errrr 87) songs and what dances I and others have matched with those songs. Check out the documentation for more details.
One fun variation on this search is to add the tag DWTS (for Dancing With The Stars) to this search. You end up with 21 songs that were both used in the last couple of seasons of DWTS and are on the Spotify Top 100 list for 2015. Another fun thing to do is to add your favorite style of dance to the search, if I added West Coast Swing to the Top 100 list, I’d end up with these 25 songs. (at least today – if other’s vote up songs as West Coast Swing, that number may change),
Do you disagree with my choice of dances for any of these songs? I would love to see what you have to say. Please feel free to comment on this post, or sign up or sign in to start tagging and voting on songs yourself.
Next year I expect that we’ll have enough activity to generate a top 100 list directly from the songs that you’ve chosen.
I know for me there are a few songs that I spent way too much time listening to while learning choreography or cutting medleys. I never want to hear them again, much less dance to them. And as much as I love music, there are some songs that I just don’t like right off the bat. So I don’t want either of these showing up time after time because other people find them to be particularly good Rumba or Foxtrot music.
Up to now that’s been a bit of a problem with the music4dance site since there wasn’t a way to explicitly like or not like a song, Everything was based on collective voting to match a song with a dance. But today I’ve added a feature where you can like and dislike a song. Then by default when you’re signed into the site, you won’t see those songs in your searches.
While I was at it, I added a few additional ways to filter searches based on how you’ve tagged and liked songs. Check them out on the Advanced Search page (remember you have to be signed into your account and have tagged or liked some songs for this to be useful).
And while you’re thinking about this, I have a question. What is a reasonable opposite of ‘like’ in this case? I’ve been using ‘not like’, but hate seems like it might be more accurate. Unlike and dislike were other options. But none of these seem quite right. If you’ve got any ideas, please comment on this post and let me know.
Music4Dance was conceived to help dancers find music that inspires them to dance. My posts about this from last year are still right on point even if the screenshots are a bit out of date. I have been adding capabilities to the music4dance advanced search control as they are suggested and as time permits. And it got a bit out of control, so to speak. As of yesterday, that control looked like this:
And now thanks to some hard work by the fabulous music4dance intern, the advanced search form looks like this:
Much improved, no? Basic documentation is available here and we’ll be updating the rest of the affected documentation shortly. We’re also looking at other advanced search features. Saving your searches and making it easier to filter on songs that you have tagged are two of our top candidates. What else would you like to see? Please let me know by leaving feedback in the comments section of this page or via our feedback form.