Zemanta launches, I'm not convinced

[#1]

Zemanta is a British/Slovenian startup which has developed a plugin for Wordpress to suggest relevant links, pictures, and tags by analysing the content of a new blog post before it is published. Zemanta was one of the 6 winners of SeedCamp last year and recently announced a $1.5M seed round so someone clearly believes in their product. I've tried the demo on their site and I've read through their blog on which the plugin is used liberally and I'm just not convinced it's useful. Take, for instance, this post on their blog. Read the text below and try to guess what the links point to without hovering your mouse cursor over the links.

Remember the post about technology today? In that same hotel, we are now accustomed to simply reset the wireless each time we pass by it, and yesterday an interesting point arose from that. If there’s anything Microsoft should patent in this world is the concept of reboot. It just works, it applies to any contemporary technological gadget, and in was in fact popularized by them.

Did you get any right? Firstly, why on earth would I want links from that article to the Wikipedia entries for technology, patent and contemporary. Secondly, bloggers have developed a style of linking relevant content to words in a paragraph which usually allows the reader to tell roughly what a link will point to without actually looking at its URL. Zemanta breaks this convention by encouraging authors to pepper each post with irrelevant links. In the opening paragraphs of this post you can guess what each link points to without visiting it. As a little bonus, the SeedCamp link points to the page detailing the 2007 winners rather than simply the SeedCamp homepage. If this were a Zemanta 'enhanced' post, the 'demo' and 'blog' links might be links to researched, considered, relevant content but, from what I've seen, they could just as easily be links to the Wikipedia entries for 'demo' and 'blog'. In their favour, I have to admit that the image suggestion is pretty useful. Images in blog posts are often only semi-relevant but serve a useful purpose in 'lightening' the page and attracting reader attention when viewed in an RSS reader. Finding such images with appropriate licensing and inserting them into WordPress can be a pain and Zemanta may be worth pursuing for this alone.

[#2]

I see what you mean with the inline links. The whole point of the linking is to provide additional context to your blog and therefore to increase its value to the reader. With Zemanta's linking, it actually can do the opposite, dragging your readers eyeballs away and preventing people from finding your post an enlightening experience. The image stuff *does* look fun though. Does it *always* return content that is allowed to be used in that context?

[#3]

Hy from Zemanta! Actually we try to be the good guys. We suggest stuff that _could_ be added to a blog, but it is up to blogger's taste to decide what and when to link to. We hope to expand the number of destinations we are linking to also. The post you picked up for reviewing is really linked way to liberally. Will fix that :) About the images. We do everything possible to inform the blogger what the restrictions on the image are and then he has to decide. And we only recommend images that should by some general standard be usable by blogger (creative commons, fair use, public domain, promotional...) And on the long run we hope to automatically learn from the blogger what he personally likes to link and use that knowledge in the engine. If you have any more ideas, please share them on your blog, via email or at getsatisfaction.com/zemanta where users can discuss the ideas about the service. Hope to hear from you again, don't give up on us! :))

[#4]

Thanks for the reply Andraz! I'll get Tam to look deeper at the plugin.

[#5]

I understand that it's still early alpha days and there's scope for the algorithm to be improved but it's more the approach that bothers me. When I read a blog I expect the links to be well researched and actually useful in the context of the discussion rather than merely on a similar topic. It seems to me that the algorithmic nature of Zemanta is encouraging bloggers to link to stuff just for the sake of it and probably without even reading the target content. If a clear distinction is made between human-selected links and computer-selected links then this becomes less of a concern. For example the 'Related articles' at the end of this post are almost entirely unrelated but at least I know that the links in the body of the post have been hand-picked and presented in an appropriate context.

[#6]

Hmmm, it might be the question about how you expect bloggers to use the service. We would expect the bloggers that want to have quality content to use the recommendations made by Zemanta as suggestions. Then checking them out and deciding if they are on topic and important for their current post. Just trying to make their work easier. Also many people find recommended images very handy, since they use them for illustration and decoration of their own blogs. If we do good suggestions they keep them, if we don't they go to flickr or some other site and find better ones. Naturally there will be the ones that just click "apply all" and never check anything, but that means that they just don't care. With or without Zemanta. We can't really improve or worsen the attitude blogger has toward his posts. :) Well I hope we can discuss this further. My question would be the following: what would you like this kind of 'automatic research' tool to provide to you as a blogger? Maybe we can find some new and interesting ideas about that? bye andraz