« Taxonomy Platforms |
| Favorite Films of 2011 »
Wednesday, February 15, 2012
Last week, I looked at commonplace books
and various implementation solutions. Ideally, I wanted something open and flexible that would also provide some degree of analysis in addition to the simple data aggregation most tools provide. I wanted something that would take into account a wide variety of sources in addition to my own writing (on this blog, for instance). Most tools provide a search capability of some kind, but I was hoping for something more advanced. Something that would make connections between data, or find similarities with something I'm currently writing.
At a first glance, Zemanta seemed like a promising candidate. It's a "content suggestion engine" specifically built for blogging and it comes pre-installed on a lot of blogging software (including Movable Type). I just had to activate it, which was pretty simple. Theoretically, it continually scans a post in progress (like this one) and provides content recommendations, ranging from simple text links defining key concepts (i.e. links to Wikipedia, IMDB, Amazon, etc...), to imagery (much of which seems to be integrated with Flickr and Wikipedia), to recommended blog posts from other folks' blogs. One of the things I thought was really neat was that I could input my own blogs, which would then give me more personalized recommendations.
Unfortunately, results so far have been mixed. There are some things I really like about Zemanta, but it's pretty clearly not the solution I'm looking for. Some assorted thoughts:
- Zemanta will only work when using the WYSIWYG Rich Text editor, which turns out to be a huge pain in the arse. I'm sure lots of people are probably fine with that, but I've been editing my blog posts in straight HTML for far too long. I suppose this is more of a hangup on my end than a problem with Zemanta, but it's definitely something I find annoying. When I write a post in WYSIWYG format, I invariably switch it back to no formatting and jump through a bunch of hoops getting the post to look like what I want.
- The recommended posts haven't been very useful so far. Some of the external choices are interesting, but so far, nothing has really helped me in writing my posts. I was really hoping that loading my blog into Zemanta would add a lot of value, but it turns out that Zemanta only really scanned my recent posts, and it sorta recommended most of them, which doesn't really help me that much. I know what I've written recently, what I was hoping for was that Zemanta would be able to point out some post I wrote in 2005 along similar lines (In my previous post on Taxonomy Platforms, I specifically referenced the titles of some of my old blog posts, but since they were old, Zemanta didn't find them and recommend them. Even more annoying, when writing this post, the Taxonomy Platforms post wasn't one of the recommended articles despite my specifically mentioning it. Update: It has it now, but it didn't seem to appear until after I'd already gone through the trouble of linking it...) It appears that Zemanta is basing all of this on my RSS feed, which makes sense, but I wish there was a way to upload my full archives, as that might make this tool a little more powerful...
- The recommendations seem to be based on a relatively simplistic algorithm. A good search engine will index data and learn associations between individual words by tracking their frequency and how close they are to other words. Zemanta doesn't seem to do that. In my previous post, I referenced famous beer author Michael Jackson. What did Zemanta recommend? Lots of pictures and articles about the musician, nothing about the beer journalist. I don't know if I'm expecting too much out of the system, but it would be nice if the software would pick up on the fact that this guy's name was showing up near lots of beer talk, with nary a reference to music. It's probably too much to hope that my specifically calling out that I was talking about "the beer critic, not the pop star" would influence the system (and indeed, my reference to "pop star" may have influenced the recommendations, despite the fact that I was trying to negate that).
- The "In-Text Links", on the other hand, seem to come in quite handy. I actually leveraged several of them in my past few posts, and they were very easy to use. Indeed, I particularly appreciated their integration with Amazon, where I could enter my associates ID, and the links that were inserted were automatically generated with my ID. This is normally a pretty intensive process involving multiple steps that has been simplified down to the press of a button. Very well done, and most of the suggestions there were very relevant.
I will probably continue to play with Zemanta, but I suspect it will be something that doesn't last much longer. It provides some value, but it's ultimately not as convenient as I'd like, and it's analysis and recommendation functions don't seem as useful as I'd like.
I've also been playing around with Evernote more and more, and I feel like that could be a useful tool, despite the fact that it doesn't really offer any sort of analysis (though it does have a simple search function). There's at least one third party, though, that seems to be positioning itself as an analysis tool that will integrate with Evernote. That tool is called Topicmarks. Unfortunately, I seem to be having some issues integrating my Evernote data with that service. At this rate, I don't know that I'll find a great tool for what I want, but it's an interesting subject, and I'm guessing it will be something that will become more and more important as time goes on. We're living in the Information Age, it seems only fair that our aggregation and analysis tools get more sophisticated.