Tag manager, tag selector, tag list in different order

In the Tag Manager the existing tags are listed in alphabetic order.
In the tag selector, they appear to be in order of creation.
And the tags listen in an album view appear in random order.
(This was inherited from Sooloos.)

Furthermore, the tag selector (which is used more often) shows the album count for each tag, which is useful but requires a single column which makes the dialog scroll.
The tag manager is more compact in two column format; I don’t know if it expands or scrolls, I haven’t filled it up.

I confess I don’t use tags a lot (beyond diagnostics and tracking issues), but clearly this needs some work.

Mind sharing some more details about how you use tags, and what would work best for you? Better sorting? Filtering? More random :slight_smile:

Aha – you asked, so you’ll get an earful!

I mostly use tags for grouping. A few examples:

  • I have a group I call “Mediterranean” that includes Anouar Brahem from Libya, Vassilis Tsabrapolous from Greece, Dino Saluzzi from Argentina and others: Dave Holland did a duo album “Hands” with a flamenco guitarist, there is some art-oriented tango from Yo-Yo Ma to the soundtrack of the movie “Tango”. It isn’t “latin”, obviously (Brahem!). It is mostly chamber jazz inspired.
  • I have a group called “Nordic avant-garde folk jazz” that includes everything by Lena Willemark, plus some albums by Anders Jormin and a few Norwegians.
  • I have another group called “Asian avant-garde folk jazz”, think Silk Road Ensemble
  • I have combined Miles Davis’s Workin’, Steamin’, Relaxin’ and Cookin’ into one group, because they came out of a two-day recording session and were broken up by the label.
  • I have a To Do group of albums that have recently been mentioned by a friend or in a review or article, things I want to listen to but haven’t had time to do yet.

They are eclectic: membership is determined by my whim.
They are not like genres or other industry classifications, there is an extra unstated filter, it’s stuff that I like.
Many of them contain artsy chamber jazz because that is my favorite kind of music. But I also have a few other groupings, like “Swedish nostalgia” (our home country) and “Swedish Christmas” (especially relevant now).
This is not a hierarchy, it is a “polyarchy”, several albums are in several groups, e.g. Trees of Light by Anders Jormin, Lena Willemark and Karin Nakagawa is in both the Nordic and Asian categories.
In some cases a tag is driven by the artist (Anouar Brahem), in some cases they are not (Dave Holland’s Hands is an exception, the rest of his work is not Mediterranean).

But tags don’t work particularly well for this purpose. For example, I have an overarching groups called “Avantgarde folk jazz” which includes everything from the Nordic and Asian groups. I can do that by selecting the tags for Nordic and Asian (logical OR), selecting all, and tagging them all with the overarching group. But that is a static assignment, when I add a new album and add it to the Nordic group it doesn’t get automatically included in the overarching group. I would like group hierarchy.

I haven’t looked at the recent improvements to Genres, I gave up early on Genres, I know they include hierarchies, maybe they are the answer.

I don’t use playlists much, because they are based on tracks and I think in terms of albums, and since playlists are designed for a specific purpose they don’t offer a very convenient browsing interface.

I think the problem is that tags are a mechanism. They have no conceptual or semantic meaning in the information architecture of the library. So the user interface is mechanism oriented, and quite primitive for any of the purposes where we might use the mechanism.

I have written a lot about grouping. Genres are industry-defined groups. A work is a composer grouping of parts. An album is sometimes a label-defined grouping of works, but in jazz and rock the album is the work, most often. A playlist is a group with special semantics. An artist forms a group over a set of albums. I want to be able to define my own groups. And I want a convenient group-based browsing interface.

A group-based browsing interface is not the same as a focus. Focus is a very rich function that can be used for a lot of things, I love it, but by its very nature and because of its richness the user interface is not fluidly integrated in the browsing experience. The popup interrupts the browsing, you define a filter, and then you get a new view and you continue browsing. Very powerful, but not fluid. (By the way, when I do use tags to aid in filtering, like you say you do and I occasionally do, they are not included in the Focus dialog. I understand the value of having the tags surfaced directly in the browser, but when I do want to combine tags with other focus mechanisms I can’t do it from one place, I have to jump out of the focus dialog and back in again.)

In general I strongly dislike popups. They interrupt the flow, selections in the popup are not reflected in real time, they are visually fragmented. Hate them. Very. But that is secondary, I think the central issue is about the conceptual model.

So if you tell me Genres are the answer, I’ll go and study that, chastened. If Genres don’t do it, we should think about a more meaningful grouping concept, and reduce the emphasis on the tag mechanism.

1 Like

Great feedback. I’m going to think about this for a bit, and come back to you. Thanks!

A few thoughts about what not to do.

In my life (I build datacenter and cloud systems management) we have this problem. Explicitly enumerating the members of a group no longer works: in a virtualized world VMs come and go, manual enumeration is both cumbersome and error prone. So we need to define rules, in effect define query-based groups. And those queries are not as simple as SQL, they often involve graph navigation, just as we see in the music world.

And they often involve a combination of graph navigation and attribute query, sometimes on idiosyncratic, non-technical attributes. It is fairly easy to find all web servers through a query. But I want to find the web servers in the front-end of the order processing system: graph. But I have several instances of the order processing system, I want to find the one for the EMEA (Europe, Middle East, Africa) region, and even there I have a production and a staging version, so this involves finding a root object through idiosyncratic query, followed by graph navigation, followed by attribute query.

But I strongly urge you not to go there. It takes a lot of system expertise to use advanced querying well, and it takes a mindset that does not go well with enjoying music. The old Sooloos example, Miles Davis albums from the 70s and 80s but not the funk, is cool but nobody ever does that kind of stuff. And if they want to, you in fact have that capability through Focus and abuse of tagging, so we don’t need to strengthen those capabilities.

The examples I mentioned are fairly simple, but I do want more intelligence than enumeration. Take the Nordic avant-garde folk grouping I mentioned. I added to it all albums by Lena Willemark, in fact all albums with Lena Willemark, but some albums by Anders Jormin.

What would be really cool would be if the system noticed this pattern and generalized it, so when I add a new album by Willemark it joins that group. But then, what to do when I add an album by Jormin? Make a guess based on the odds of Jormin being in the group? Maybe; it will be error prone, but the errors are not fatal, and they are easy to correct. Ask the user? No no no, I hate machines prompting me, the machines don’t tell me what to do, I tell the machines what to do.

But nonetheless, I think this is an interesting direction to go. Systems are emerging with the level of intelligence to do a good job on this kind of pattern matching, from Outlook’s “Clutter” feature (intelligent junk mail filtering) to Google’s advertising. The point is, these systems are cloud-based, both for computer resources and for the large data volumes (cross-user) to form robust conclusions, and they are complex to build.

So if we can’t go there right now (you have other urgent things to do), it is nonetheless interesting to consider this as a future direction. (I have mentioned other benefits of cloud processing of cross-user metadata to @Danny.) And if we think of this as the north star, you can then think about the near term direction as a step in that direction. It may help you choose approaches conceptually in line with this direction, and avoid those that walk elsewhere.

It’s an interesting problem.

Have a happy holiday. I for one intend to enjoy some music.