As the year gets to the end, everyone writes “Best of” lists for the past year. I thought I would do something similar, but since we’re at the end of not only a year, but a decade, it’s worthwhile to reflect on the changes in how people manage and organize their increasingly digital stores of information. Over the next week, I’ll highlight some major developments and discuss how they’ve informed the development of Mendeley. This week it’s the practice of tagging bits of information as opposed to filing things in a hierarchical folder structure, with posts on the move to querying databases of information as opposed to loading information from individual files and the representation of information as a temporal stream as opposed to a static page to come next week.
Tagging vs. hierarchical folders
Websites were once composed of individual files which were arranged in the structure of the website and linked together. Around 2000, a new breed of blogs and web services became popular, which took a different approach. When you visited the site, the main content of the page was dynamically fetched from a database and used to construct a page customized to you. Because these pages were database driven, they were more dynamic – the page you saw changed as you interacted with it or as information was added. Furthering this trend, they began to collect all sorts of information about the information they were storing.
The prototypical example in this case was a little service called del.icio.us*. A pioneer of the Web 2.0 movement, del.icio.us allowed you to save links, not to the bookmarks folder of your browser, but to their webservice. Instead of filing the bookmarks in folders, you gave them a couple of descriptive tags. You could then use those tags to get the service to show you a page constructed using those tags. You could just go to http://del.icio.us/tag/whatever+whatever and get a page that showed all the tags in the database that had been uploaded by anyone. It was a fascinating use of collective intelligence and one of the first popularizations of the concept of tagging. It’s also where many academics first understood what this whole Web 2.0 thing was all about. Del.icio.us showed that for the same effort it would take to file a link in a folder structure, you could tag the link instead and get back so much more than you put in.
Tagging instead of filing in hierarchical folders has other advantages as well. With tags, you don’t have to pick an organization scheme up front. You don’t have to decide if you’re going to file according to the organism the research was done in, or by the disease, or by the cell type used. You just tag and move on. You’ll never run into the problem where you realize that one folder is getting too big and should be split into subfolders because each tag works like a separate folder. If you want Mendeley to show you all the files tagged with “Wnt” and “bone” just type “tag:bone,Wnt” in the search. This gives results identical to filing the papers in a Wnt folder and a bone subfolder, except you don’t have to decide which is the top level and which is the subfolder – you just tag and move on. Like the original delicious tags, you get results just like with folders, except you don’t have to set anything up before hand, and you get back so much more. Having tagged your papers, you can go to Mendeley Web and search by tag, finding all the papers everyone has tagged with that term. Tagging is easier to get started with, and it’s easier to maintain.
Of course, for more lightweight use, it’s sometimes easier to just click on a folder and see what’s in it. Folders aren’t going anywhere in Mendeley and they work kinda like tags in that you can have a paper in multiple folders, but you lose the neat search abilities that you get with tags. Subfolders are currently the 4th most requested feature on our Feedback Forum, and we’re listening to our users about this, but I hope some of you take a moment to consider if tags will work for you.
2011-5-18: Fixed typo in multiple tag search syntax
Mendeley Blog 
A downside to tagging seems to me how to ensure what I believe is called ‘integríty’ in database terms: using the same tags when appropriate and avoiding misspelt tags. Or see Alex Payne’s case against ‘everything buckets’: http://al3x.net/2009/01/31/against-everything-buckets.html, argueing that computers work best with structured data.
In light of this article, are we likely to see tag-based methods make their way into the evolving API (http://dev.mendeley.com/)?
To make people use tags instead of folders in Mendeley, allow more control over tags, like show the tag cloud and allow to (batch) edit tags (to correct tags or group them, as you have it already for authors, journal names, etc.).
querying databases of information as opposed to loading information from individual files
Technically, there is not a big difference there. A single file can be present in multiple folders, and, at least under Windows, the hard drive is indexed similarly to the way databases are — with a B-tree.
I disagree with your arguments. There, I said it!
1) We know Mendeley team has had difficulties implementing the hierarchical folders in their web interface. I believe that is the reason why this structure hasn’t yet been implemented in the desktop version as well. So, it seems to me that by this article someone has tried to simply justify their shortcoming rather than actually addressing them.
2) For the sake of argument: tagging is not the “modern” way (as in, hierarchical structures are “obsolete”). Tagging is the just newer than hierarchical structures. There’s a big difference. So let’s not get carried away. Hierarchical structures have been around for a long time and for a good reason — they are the “natural” way of organizing things. Human brain works by categorizing information. We almost always compartmentalize and categorize or else we’d go mad with flood of information and the chaos it inflicts. Tagging, on the other hand, is simply an additional mean of retrieving information. It’s nice but it’s complementary to hierarchical structures not a replacement. Hope this makes sense.
I guess I’m missing the point. How are tags different from keywords? This seems a bit redundant to me. Most of the metadata that we get from other databases and load into Mendeley already has keywords. Why re-invent this?
Alec – Great question. The difference between keywords that come from other databases and tags is that the keywords are set by the author or publisher and are often simply extracted from the document. Tags are supplied by the person storing the document, which means that it reflects what the document means to them, not to the publisher or author. It’s just a different kind of metadata, useful because they often contain words or concepts that aren’t present in the document, such as how the document is related to others the person has in their personal collection.
Dara – You’re exactly right that tagging and hierarchical structure can work side by side. I’m not trying to start a fight here (well…maybe) but I would argue that neither is more “natural” than the other. Labeling things and opposed to putting things in boxes are both ways of organizing something. As I discuss above, for smaller collections boxes aka folders work just fine. However, as the size of the collection grows, you often find yourself needing to separate one box into a couple boxes, or move things from one box to the other. Tags avoid that issue entirely. That’s all I’m trying to say – just introducing the concept of tags as an alternative means of organizing information, one which works particularly well with our system.
Apart from the discussion of tagging vs filing, I think the current implementation in Mendeley is kind of incomplete.
Right now, you can add tags to documents and groups, but it is hard to effectively retrieve the data, so it kind of feels like a one-way system.
What makes Delicious so good is that it is intelligent about tags. It remembers the tags you and others assigned to pages and presents these as “suggested tags” while bookmarking. On your own profile page, you can subsequently see all the tags you ever assigned, aggregated by the number of pages you assigned it to. This allows you to use the tags as keys to the items you assigned them to. I believe these things are currently lacking in Mendeley’s interface.
First of all, there should be help while tagging. A free-form text input field simply doesn’t cut it. My memory of tags I used previously is not that good, and typing or spelling errors mess it up. Secondly, I feel a lot can be improved to the way tags are used to access documents. After some looking, I found the “Filter by My Tags” in the drop-down on the bottom left, but this is kind of limited. This list should at least show counts for the tags. My preferred solution would be to have something like “saved searches” in the library bar on the left (where groups and trash are).
Entirely agree, Michel. Improvements to the design of the tagging system is coming, including tag auto-complete, tag counts, and more. Just to make sure you know, if you do misspell a tag and want to correct it, just drag and drop the wrong tag onto the right one to rename all instances.
William – i like your argument that tagging can be just as (or more) effective than heirarchical folders, but the functionality doesn’t appear to be working in Mendeley as you described it. The only way I can get Mendeley to perform one search for files that have two separate tags (e.g., “Wnt” and “bone” as you described) is using this syntax: tag:Wnt,bone Any use of spaces screws up the search for tags.
Also, the document tagging functionality can’t handle multiple word tags. For example, if I tag a document with a tag “landscape genetics”, it’s not different than using two tags “landscape” and “genetics”
You’re right, Cameron, there should be a comma there. For multi-word tags, use quotes: tag:”landscape genetics”