Building a better network: Identifying trends/posts of interest

When you build a network of blogging networks, the problem quickly escalates from “how do I collect as much data as possible?” to “how do I manage all this data?”

Take a look at the Science Blogging Aggregated home page. There’s lots of great stuff there — too much for the typical reader to handle. Even if you visit several times a day, the information rushes by too quickly to discern any trends, and it’s hard to know which posts are really well thought out and which are just one-off posts that hardly merit your attention at all.

We talked yesterday about one way of sorting through the data — tags. However, this method alone probably won’t satisfy all users. A person might be interested in all posts tagged “psychology,” but they might just want to see the highlights of what’s going on in other fields, and tagging won’t help them identify the most interesting, thoughtful posts.

We see at least four possible ways of sifting through the posts to find the most interesting ones.

1. Crowd-sourced ranking. Users rate or recommend posts they like, so others can sort by rating or number of recommendations to find the posts they want to see. An advantage is that there is no central authority telling readers what to like. A disadvantage is that blogs that are already very popular are perhaps most likely to be recommended, so this system might not help users identify up-and-coming blogs that are very high quality.

2. Self-promotion. Bloggers could promote a small number of their posts, indicating these are their best work (one per week? one per month?). This overcomes the “up-and-comer” problem, but a blogger whose work is mediocre could exploit the system by promoting posts that aren’t very interesting or useful to others.

3. Active curation. Editors could be chosen for each field (physics, biology, etc.) and actively promote one or two posts each day. That way readers would know that an expert has read all the posts on a topic and selected the most interesting or relevant. Advantages are that editors may be able to identify trends that more automated systems don’t catch, and that editors may be less swayed by the most popular blogs. Disadvantages include possible bias of editors, and variable editor quality. It would also require coming up with a system for selecting editors. Would a central person be in charge of that, or would we need to create some sort of a system for nominating/voting for editors?

4. Social networking. We could create a truly social network where users are only shown the “likes” of their friends. However, this requires a significant programming effort, and people are reluctant to join new social networks when they already participate actively in one or more networks. I think we might be better off using the social features of other networks, rather than building our own. If we could make it really easy for people to post their “likes” to Twitter and Facebook, then we could leverage those networks to perform the social function.

There is, of course, no reason that we shouldn’t do all of these things over the long run. But we have limited resources. Which of these approaches is most useful? Are there any other approaches that would work better? Do you have any specific suggestions for how to implement any of these ideas? Let us know in the comments.

3 comments on “Building a better network: Identifying trends/posts of interest

  1. Pascale says:

    I use twitter to filter for me. If my twit-buddies like it, I know it’s worth a look. If I find myself repeatedly going to the same blog, I add that person to my follow list.

  2. Stacey says:

    I have been struggling with this myself. I use Twitter to some extent, and can weed out some of the more useful stuff that way. But I definitely need a better strategy to sift through everything I am interested in. More time would help! Thanks for the post.

  3. Don Sawtelle says:

    Along with the possible “sifting” methods you mention, scienceblogging.org could help readers find posts on topics they’re interested in by using an automated classifier.

    We just now launched http://winnowtag.org to illustrate how a classifier can address the problem of an overwhelming volume of information. winnowtag.org can find on-topic posts in a volume of material so large that typical methods of organization are impractical. It uses Winnow, a high-performance text tagger we’re releasing as open source.

    There are various ways scienceblogging.org could integrate the Winnow classifier. A simple example: for each field in which an editor identifies the best posts, Winnow could use those posts as examples in order to automatically find other posts on the same topics.

    Winnow does work surprisingly well. winnowTag.org downloads and tags 7,500 feeds daily and keeps the items for three months, thus currently has 683,000 items on a bewildering variety of topics. Here are a couple of illustrative tags:

    entomology – http://winnowtag.org/#mode=all&tag_ids=468
    space – http://winnowtag.org/#mode=all&tag_ids=11

    A higher number at the left of an item means the classifier is more certain that item is a correct match for the selected tag.

    http://doc.winnowtag.org has more info. We’d love to have your feedback. And if you did try integrating the Winnow classifier, we can give you support.

Leave a Reply