Library clips

The other day on my Grazing Lists post I mentioned reading feeds by relevancy.

Reading Behaviour

Relevancy can be based on your reading behaviour…what you click on, see Findory Favourites

…also other reading behaviour such as stuff you save, share, email, etc…since you can do all this in Rojo, they can gather all this as attention data, besides just what you click on.

Popularity

Another way you can work out relevancy is by incoming links…you can say organise the posts in my RSS Reader by order of popularity, that is, the post that has the most incoming links from the blogosphere will be first and so on.

Or even the post that has the most incoming links from blogs within my RSS reader (my mini blogoshere) will be first and so on.

Or, besides being based on just your RSS reader (OPML) you could also extend this to other OPML’s of your choosing.

But is this personal relevancy, sure you have the feed set you want, but you want to read the relevant posts first, but this is based on what everyone collectively likes, not what you personally like…so I really think this method alone is not relevancy, it is just link popularity…it only goes so far.

So maybe combining reading behaviour and popularity will make a more robust relevancy, ie. you will see posts you like on the top of your pile, based on what you read in the past, but you will also see posts on the top of the pile based on what everyone likes (so this adds a bit of recommendation within your own feed set…it’s stuff you might read anyway as you get to the bottom of the pile, but it is recommending to push it up to the top…so we are not recommending posts from other feeds that you don’t subscribe to, it is recommending to only push stuff up that you already have).

Threading

Another subtle form of relevancy, actually a second level type of relevancy, is that when you are reading your feeds (either by date or relevancy), the current post your reading may have links to other websites (as they usually do, eg. everytime I mention Technorati in a post I hyperlink the word to the homepage).
Now if the post you are reading has a hyperlink to the Technorati homepage, the RSS Reader will search through all archived posts and current unread posts, and look if any of these posts have a hyperlink to the Technorati homepage…if it does find some it will list these in your current post.

Check out SharpReader for this feature.

Instead of threading these common hyperlinks, Rojo has decided to aggregate them all in a list, they call it Recommended Links.
Basically you can see a list of all the webpages that the posts in your RSS reader are pointing to, if 20 posts in your RSS reader point to the Technorati homepage, then this may be on the top of the list.
In the end, this is a type of recommendation system, based on what you’re feeds are pointing to, not on your reading behaviour…it recommends websites based on your feed set.

Chuquet has made this into a service, the future plan is to enter your feed set (OPML) and it will scan all the hyperlinks in each post and then display them in order of which website got linked to the most (again, popularity)…based on time.

TailRank takes into consideration linking behavior, the text off the post, links in common with other users, search relevance, and various other factors for recommendations.
So this is more a recommendation for posts not in your RSS reader, rather than ranking the posts in your RSS reader by relevancy (I think?)…not too clear what the filter means.

Megite seems to organise posts in your OPML by popularity and also threads other common posts…hmmm sounds the most interesting so far.

Actually, it doesn’t list all the posts in your OPML in order of relevancy, what it does is list the most popular items each hour (or some time range)…eg. so instead of reading 100 posts a day, you are reading maybe 10 posts per day. It will also thread related items, ie. posts that are about the same/similar topic (maybe this is based on text analysis…this is what you see on Google News, and Memeorandum, not sure if Waypath Related works this way, it seems to).
For each item, and also the related items it lists incoming links (maybe these are generated from Technorati, I’m not sure).

Not sure if the threaded posts are the incoming links to the featured post or are just related because they have common links, or common chunks of text…or maybe a combination of all of these.

NOTE: I’d also like to see threaded posts based on text analysis…if a past or unread post in your RSS reader has lots of common words to the current post you are reading maybe these posts can be linked to the current post you are reading, as well.

You will also see some posts that are not from feeds in your OPML, these are thrown in for good measure…maybe it incorporates a bit of what Chuquet does (if a lot of your posts in your OPML are pointing to a website, this website might be relevant to you, so it is made more noticable).

But it can’t track my reading behaviour, this stage of relevancy can take place at my RSS Reader (this is the relevancy tools of Findory, and Rojo)

…this means I could load an OPML into Megite, then grab the feed, and just read this one feed daily.

I still want to read all posts from my OPML, but just by a relevance like popularity…

See more:
RSS Reading: recommendations
RSS Reading: different views

Back to relevancy via incoming links

I’d like to see other interaction besides just incoming links (we have already talked about reading behaviour), but what about comments, if a blog has lots of comments, maybe this could be a popularity measure as well, etc…see more and more.

And maybe the quality of incoming links can be taken into account, if one of the incoming link to a post is from a post or blog that has lots of incoming links itself, this could boost the popularity of the post.

Anyway, as we have mentioned linking activity is not really personal relevancy, it is based on the collective popularity.

A perfect example of this is the OPML Sampler…order posts in your OPML by popularity (incoming links).
This only shows the popular posts from these feeds, all other posts are discarded, so there is only the top of the pile…this is a bit risky, I would like to see all posts, even if I don’t get to read stuff from the bottom of my pile (way down the river), I’d still like to know it’s there if time permits.

So megite seems to be an enhanced version of the OPML Sampler, beside the quality of results, it also clusters/threads items.

Risks

The risk of using megite is obviously that you are not going to read every post in your OPML, you will just see the popular posts, if you are happy to rely on this, good luck, I’m scared I’d miss out on stuff…although it’s great for a quick 5 minute browse if you haven’t got time to read your RSS Reader.

What is needed is another view, where it lists all the posts in your OPML based on popularity, that is, all the posts, not just the top posts.

SCENARIO : reading all posts in your OPML based on popularity

Imagine you are reading posts in your RSS Reader by popularity (based on incoming links).

Towards the top of your river of news are posts that are considered popular in the blogosphere (not necessarily popular to you)…as you read through your posts it slowly becomes apparent that they are becoming less relevant…wow, this is great, it works, less popular = less relevancy for you (personally I don’t think this will happen).

Anyway you decide to mark that feed or river of news as read without reading all the posts, as they are starting to be less relevant…logic tells you, stuff on the bottom of the pile won’t be relevant at all, so you mark the feed, or folder as read, confident you are not missing out on relevant stuff.

But what if you waited a few more days till you read your RSS reader, this would of given some posts towards the bottom of the pile more time to swim to the top, as there is more time to gain some incoming links…so if you waited, a post could of reached near the top that may of been relevant, but because you read your feeds a couple of days earlier, this post was at the bottom of the pile, so you didn’t even see it, you marked “as read” not bothering to read stuff from the bottom of the sea.

An option is to never mark your feed or folder “as read”, so all posts are still there except the one’s you clicked on, this wouldn’t work on Bloglines, but it would work on Rojo, Attensa, etc…ie. when you click on a post in Rojo or Attensa, it is considered read, whereas in Bloglines you can’t mark a single post “as read”…Attensa is even more useful as you can mark a single item back to “unread”.

And what about those feeds that aren’t popular or just starting, they may never have many comments or incoming links, but still they may be relevant, so how will these float to the top.
Maybe you can have a popularity slider, a la Technorati, or read your relevancy river of news in reverse.

Another idea is to have some overide options on your relevancy, like, show me all posts that have the words opml or “reading list” in the title or the body first, then show the rest as per usual…or automatically delete any of these terms or show them as least relevant.
Bloglines can do at least “search for” feeds, but I don’t want to generate a feed, I just want this as part of my relevancy.

megite for my RSS reader

I’d like to read all posts in my RSS reader by popularity (like megite), but I’ll stress again, all posts, not just the top posts.

For each post the incoming links are listed
(based just on your OPML, or a bunch of OPML’s, or the whole blogosphere)

For each post, there are related items, based on text analysis,
(based just on your OPML, or a bunch of OPML’s, or the whole blogosphere)
- incoming links are also shown for these items

For each post, common links are threaded
(based just on your OPML, or a bunch of OPML’s, or the whole blogosphere)

I can manually override the relevancy, to push certain posts to the top of the pile, or bottom of the pile based on keywords

Popularity slider, and read posts in reverse.

What if your RSS reader based relevancy not only on popularity, but also reading behaviour, or you had 2 choices of relevancing reading, and a 3rd combined choice…Rojo is half way there.

RSS reader: database queries

I can use the data to make time range queries

Example of reading by relevancy
- show me all posts (by popularity)
- show me posts in the last day (by popularity)
- show me posts in the last hour (by popularity)
(this would make it more like megite)
- show me posts in the last month with the keyword “reading list” (by popularity)

Memetrackers

Check out Steve Rubel’s post on memetracker wars…tracking the word Breememe…see the results (the comments on both posts clarify the situation).

Search engines will show every occurrence of a term, link searches (eg. Technorati) will show every link to a post, what a memetracker attempts to do is filter out the noise, and only show you a handful of the important incoming links (what’s important - is it from the most popular blogs?)…but it will also have a link to see all the incoming links at Technorati for more.

I’m not sure of this, but what it may also do is find related items based on text analysis (common words)…actually maybe this is how they choose which incoming links to cluster.

Anyway, the idea is to give you a summary of a meme’s spread, if you have time, do a link search at Technorati for the full picture.

The 3 main players are Memeorandum, TailRank, and Megite, they all seem to perform well.

TailRank allows you to personalise this to your OPML (filter view)…although I see a lot of posts from feeds not in my OPML…what’s going on here?

Megite also allows this and all posts are from my OPML, so it does what it says…TailRank likes to include a lot more recommendations for you, a lot of this is based on other OPML’s that have common feeds as your OPML.

See more.