RSS
Email
@steve_e
Are Likes Poised to Replace Links as the Web's Primary Signal? tip @techmeme http://j.mp/hPZaVE 
I was drawn to a tweet by Steve Rubel, from Edelman Digital, earlier:
Steve Rubel
steverubel
The tweet sucked me in and so I read the blog post (great reading!).
It got me thinking: I prefer a web of sentiment than a web of likes. Likes are only one signal, and they are in themselves a sign of sentiment (cc @steverubel) .@steverubel is right, it'll take time to move from a web of links to a web of sentiment. There's so many sectors that aren't social yet But I think a web of sentiment is coming, it's just a matter of time (and someone building a sentiment index)
Steve Evans
steve_eSteve Evans
steve_eSteve Evans
steve_e
Would it ever be possible to create a sentiment index capable of measuring the value or popularity of a web page based on multiple signal factors.
There are so many factors to consider. Not only are there Links (in the traditional sense) and Likes (in the Facebook sense) there’s also Tweets and Retweets, Diggs, bookmarks on Delicious or Diigo and semantic relevance (the actual relevance of the textual, or other, content) of course. Semantic relevance in fact is key and one of the problems with search and SEO in general is that relevance has been gamed to death by Links and black-hat SEO practices. When we’re trying to discover content relevance is possibly the most important signal, even more so (or at least as important) than how popular a web page is in a lot of cases.
If I’m really trying to discover the best content I can though there is another signal which can act as a really powerful sign of positive sentiment, actual user engagement with a piece of content. High engagement with a web page suggests people like it and find it interesting or useful, to me that’s a powerful signal that indicates I might want to read it or interact with it too.
If you could combine semantic relevance with Links with popularity (from Likes etc) with actual level of user engagement you could have a very powerful sentiment based approach to discovering content relevance. I love the way Steve Rubel says in his blog post that with Likes ‘this new network of signals allows content to find you, rather than you having to go find it’. That’s the key to good information discovery, reversing the trend from being the content hunter to being hunted by content that is relevant, popular and (in my opinion) has high levels of engagement.
Content which is Liked a lot isn’t necessarily relevant, linked to a lot or engaging…
How would you measure engagement across the whole web? I hear you say. Good question. Obviously most web pages and pieces of content have analytics which have data that suggests a level of engagement from its users, but that’s never going to be opened up to make it a useful signal for information discovery (or maybe Google could use Analytics data?). Could a company like Hitwise (provider of analysis and data on website usage), who has access to web users click stream data at the ISP level, use data on engagement to create a new form of signal to add to the mix? Unlikely as it’s not their business model and also privacy laws could get in the way.
I don’t know the answer but I have had fun thinking about it today. Love to hear your thoughts! Do you think we’re moving to a web of Likes (we may well be in the near term) or are we more likely to move to a web of Sentiment where multiple signals of relevance, usage, popularity and actual user engagement are key?
Photo by retinafunk on Flickr.
By Recording Studio Will 01/03/2011 - 9:07 pm
I certainly agree Google should use analytics data – not just no of visits but look at how long users stay and how many pages they browse.
If a user left a comment (and it could be recorded in analytics) and they had stayed on the site quite a while that could be a great barometer in measuring engagement.
By Ellie K 02/05/2011 - 7:26 pm
The folks at SAS as well as IBM, Microsoft Research and academia have been trying to design effective sentiment analysis metrics for awhile. Based on SAS blog posts, as well as a research video from Microsoft that I saw recently, it is none too easy to effectively capture and quantify online user sentiment. Surprisingly, this seems to be the case despite having all sorts of computational linguistics expertise, very high end analytical software applications (like SAS or IBM’s SPSS) and lots of processing power.
I would post links, but I’m not certain that that is allowed on your site. There is no notice saying “HTML allowed for style”. Whenever I see that phrase, it makes me think, “Hmmm, my usage of markup actually has nothing to do with my sense of style, or lack thereof”! But I digress.
Actually, Microsoft’s new media player is really amazing (that’s how I viewed their sentiment analysis video). You can find it on the Microsoft Research web site. Yahoo! also has an excellent media player with nice browser compatibility support and other features. Unfortunately neither Yahoo’s nor Microsoft’s players are likely to get much attention. Usage frequency establishes format standards, for better or worse. But that’s the way of the world.
Your blog is pleasant reading. I enjoyed the airplane travel photos and links in particular. I do wonder why you chose a domain name with a dash in it: steve-e as I thought that doing so was an SEO no-no? But on those matters I have no knowledge, as I’m a statistician and risk management type.
Good luck with your business endeavors. I found my way here from the Artemis site about catastrophic risk and re-insurance. Now THAT was an excellent site name, Artemis!