The PageRank of Personal Relevancy
Ilya recently posted some great data and analysis on the current state of RSS reading habits.
He goes on to explain many of the issues I alluded to in a previous post titled "Show me the money (or the pain)".
I think the question of information overload is answered. Yes there is an overload. But RSS is not the problem. In fact blogs and user generated news are not the problem either. They are just one source of information in our lives.
There are application events, presence changes from our friends, internal memos from head office, applications on our desktop and more all clamoring for our time...
So tools that try to cluster and suggest content from blogs and mainstream news sites are only (very) useful for part of the time.
Ilya goes on to make a great suggestion in his post. He recognizes that collaborative filtering has limitations, Keyword filtering is 'so 5 years ago' and that any one 'community voting' measurement will fall short.
With Touchstone we have gone to great lengths to cover all these usage scenarios. We have built a platform that accepts 'items' not 'RSS'. This means that we can source content from places other than RSS and then cache, rank and route them in a unified way.
Our 'rank' is not based on collaborative filtering or keyword filtering or community voting or previous reading behavior. It is based on some and none of these things at the same time. As such, our technology can work in a vacuum on a personal item behind the firewall, just as it can work on a news item that the whole world can see and link (read:vote) to.
Also, there is no 'handshake' period where our application tries to track your reading behavior over time. We are on the client side which means we have access to your browser history/cache, documents and email for an instant, broad and ongoing base of 'Attention Data' in order to determine your interests.
Ilya rightly compares this to PageRank. While PageRank uses incoming links as a vote to measure authority, it relies on a broader set of factors to make a decision and produce a number.
And because the result is a number rather than a binary 'yes or no' filter or an opaque recommendation, Touchstone can make intelligent presentation decisions when displaying the alert/content/information. The bigger the number the bigger the item on the page.
My friend Adam also posted about the 'Feed Overload Problem' on his blog.
He goes on to explain many of the issues I alluded to in a previous post titled "Show me the money (or the pain)".
I think the question of information overload is answered. Yes there is an overload. But RSS is not the problem. In fact blogs and user generated news are not the problem either. They are just one source of information in our lives.
There are application events, presence changes from our friends, internal memos from head office, applications on our desktop and more all clamoring for our time...
So tools that try to cluster and suggest content from blogs and mainstream news sites are only (very) useful for part of the time.
Ilya goes on to make a great suggestion in his post. He recognizes that collaborative filtering has limitations, Keyword filtering is 'so 5 years ago' and that any one 'community voting' measurement will fall short.
With Touchstone we have gone to great lengths to cover all these usage scenarios. We have built a platform that accepts 'items' not 'RSS'. This means that we can source content from places other than RSS and then cache, rank and route them in a unified way.
Our 'rank' is not based on collaborative filtering or keyword filtering or community voting or previous reading behavior. It is based on some and none of these things at the same time. As such, our technology can work in a vacuum on a personal item behind the firewall, just as it can work on a news item that the whole world can see and link (read:vote) to.
Also, there is no 'handshake' period where our application tries to track your reading behavior over time. We are on the client side which means we have access to your browser history/cache, documents and email for an instant, broad and ongoing base of 'Attention Data' in order to determine your interests.
Ilya rightly compares this to PageRank. While PageRank uses incoming links as a vote to measure authority, it relies on a broader set of factors to make a decision and produce a number.
And because the result is a number rather than a binary 'yes or no' filter or an opaque recommendation, Touchstone can make intelligent presentation decisions when displaying the alert/content/information. The bigger the number the bigger the item on the page.
My friend Adam also posted about the 'Feed Overload Problem' on his blog.
Labels: attention, collaborative filtering, filtering, information overload, personal relevancy, recommendations, RSS, voting
0 Comments:
Post a Comment
<< Home