Clippr.
OK so google have announced Google Reader, which is an RSS client for your browser. The best way to introduce Clippr is as an RSS reader i built that shows up Google Reader in most departments. It feels good to be ahead of the google this time ;)
I built Clippr as part of my dissertation over at Imperial. It's social software for the web. I use it everyday - it's become one of the most important apps in my browser, which i guess is enough reason to tell you a little more about it.
First off, a one-line description of Clippr:
Social bookmarking meets RSS feeds. A great way to discover, archive, create and share articles online.
OK, now for the details.
Overview
Clippr bears some similarity to the social bookmarking tool Del.icio.us, in that it allows users to tag and archive content from the web. However, where del.icio.us revolves around the concept of the 'bookmark', Clippr's main building block is a 'clipping'. A clipping is either a textual excerpt taken from a webpage or an article from an RSS feed. Both these types of clippings are treated in the same way by Clippr.
Why would a user want to take clippings?
Clippings are like a fluid bookmarking strategy for the web 2.0. They're a replacement of the bloated 'temp' folder which sits in everyone's browser bookmarks. They acknowledge the kind of transitory interest in numerous topics that the web encourages in users, as well as the fact that users are increasingly browsing by RSS feed. Clippings are a replacement of the bookmark per sé, in that they are represented as RSS articles and not simply as hyperlinks. Put simply,
Clippings tie a user's bookmarking activity to their RSS reading habits and provide a richer definition for archiving web content than a bookmark.
What's the big deal about folksonomy? It's just labels.
Um, yeah, like what's the big deal about computers, they're just big calculators. OK.
Folksonomy is a faceted, emergent classification strategy. It's the natural classification strategy for networked media, in that it addresses directly the fact that,
In an information retrieval system, there are at least two, and possibly many more vocabularies present. [14]
One of the main advantages of Folksonomy is that,
Since the organizers of the information are usually its primary users, folksonomy produces results that reflect more accurately the population’s conceptual model of the information. [35]
For more on folksonomy - and the sources of these references - check my dissertation.
What's implicit social software?
Social Software can perhaps be summed up as software that recognises the most important component of it's system is the user-base. If social software wants to do smart stuff, it does so by aggregating the intelligence of it's user-base, in terms of the data captured by the application during usage.
Implicit social software means getting rid of user profiles and user visibility in general and just letting the app use the aggregated information to work out relations between data. Like this the application functions like a single-user desktop application without all the distractions of social networking, but draws on community intelligence to help the user navigate the data architecture created by the user-base.
Text-analysis is rubbish, it gives me google adsense ads about detergent when i'm reading about the SOAP protocol
True. Clippr acknowledges the user-base as the central component in it's classification strategy. The system performs text analysis on incoming articles from user subscribed feeds, which it uses to extract keywords and top stories from the fresh crop of articles. Then, whenever a user's tag coincides with a keyword extracted by the system, all articles with that keyword are shifted over into the tag-space. This is a crude automated classification strategy which reconciles text analysis and tagging - it works pretty good.
Moreover, Clippr analyses tag intersections to recognise related tags, and presents these to the user as they search for articles. It's more subtle than del.icio.us in this respect, in that it quantifies the similarity between tags, in terms of intersection magnitudes.
So text analysis needs to be combined with context analysis to be effective at picking out similarity between documents, and that's what Clippr does, by letting users provide contextual information and analysing it.
Folksonomy is an uncontrolled mess. Jack says 'web', whilst Jill says 'internet'
True. Wordnet can be used to reduce the tag space by recognising synonyms, hypernyms, hyponyms and compound terms, but it has limited performance with a completely uncontrolled vocabulary. Acronyms and neologisms abound in tagging. A lot of this is still unresolved and impacts the performance of search in folksonomy.
Show me the money
Here's the feature list:
- OPML import/export of feed subscriptions. Folders are flattened to tags and imported automatically into Clippr.
- Firefox plug-in and bookmarklet to facilitate clipping stuff from your web browser.
- Tag clippings, tag feeds, tag like a demon.
- A community oriented article base formed through user subscriptions, refreshed periodically. Full text RSS/RDF/ATOM formatted feeds are supported.
- Text analysis (article clustering) on incoming articles, in order to extract Top Stories and article keywords.
- Context analysis (tag clustering) used to recognise related tags.
- Change your tags whenever you want. Clippr handles merging/splitting of tag-spaces.
- Power editing using batch actions thanks to a gmail style dynamic dropdown.
- Tagging combined with keyword extraction to produce automated classification of articles. Text analysis and folksonomy reconciled.
- A search engine supporting a query syntax for folksonomy - search by tag (intersection/union), feed, keyword or any combination of these. Implemented as Live Search for desktop style responsiveness (it behaves like Apple Mail search - wipe the search field and return to where you were)
- RIS export for using web references in bibliographies
- RSS export of your Clippings archive.
- Mail an article to a friend or recommend it to a fellow Clippr user.
Ok, let's try it out
You can't. It's not public right now because it needs a dedicated server to run and i can't afford to run and maintain one. If you want to help me setting one up, please get in touch with me.
More information
Here's a quick presentation and here is my dissertation in full. Source code (ruby/javascript) to come.