So what would happen if there would be a reputation system which allowed the individuals who take part in social search to rate individuals within their social context based on the benefits that the results yield? The reputation score would be visible to all other individuals in this network, so reputation could be a major incentive to provide higher quality sources and to tag these sources more specificly. Therefore the social search engine should provide mechanisms to take into account reputation while ranking search results. This simply means sources added by individuals with higher reputation should generally rank higher.

We already had some hot discussions about how to introduce such a system to SLENSE and we will follow that idea for sure. Another hot question is what kind of parameters could determine reputation in social search, besides the benefits of a search result for an individual.

A gene in this framework is a particular answer to the key questions associated with a single task in collective intelligence systems (namely Who, Why, What and How). The paper presents different categories of answers for each of those four questions. For instance, for “Who” the categories are “Crowd” (i.e. anyone in a large group who chooses to participate) or “Hierarchy” (i.e. groups including someone with the authority to assign tasks to certain group members). For the “What”, the categories are “Create” (i.e. producing something collectively) or “Decide” (i.e. evaluating and selecting alternatives). Please have a look at the paper for a detailed treatise.

Applying this framework to Slense, we come to the following (preliminary:-)) conclusion:

• The answer to “What” is “Decide”: Slense allows to collectively decide what’s interesting and important on the web
• The answer to “Who” is “Crowd”: everyone using Slense has the same rights, there’s no one able to impose anything on anyone
• The answer to “Why” is an interesting one: depending on what you use Slense for it’s either “Glory” or “Money” – this is especially interesting when you consider using Slense for your business
• The answer to “How” is “Independent” and “Collection” since Slense collects the tastes and preferences of different users and allows for sharing and combining them

It’s interesting to think about Slense in that regard– let us know what you think!

Mahout currently includes a whole lot of different clustering algorithms such as k-means, EM etc. plus an algorithm to perform singular value decomposition. Other algorithms for dimensionality reduction and matrix decomposition (e.g. principal component analysis) are planned but seem to be missing yet. When it comes to classification, there is only a decision tree algorithm available right now, while neural networks, support vector machines, regression methods will be available in the future. Mahout also includes recommender algorithms via the Taste framework.

The biggest problems with mahout right now is the lack of comprehensive documentation. So we decided to put a couple of links that we found during our research on this blog:

• You can find some news on the state of mahout, hadoop etc. here
• This slide deck provides some interesting supplemental information that you can use in conjunction with the tutorials on the mahout website.
• Here is a direct link to the Taste subproject providing recommendation algorithms
• Finally, here you can find some more code and some background info at metaoptimize

Now, as it seems, this concept can also be applied to teams of so called “knowledge workers”. Imagine a team of people, recurrently researching on the web around one or more specific content areas, a classical situation today in any analyst’s, product manager’s, researcher’s office.

Many sources are being screened daily in these offices, relevant and trustworthy content is being separated from scam and the information relevant for the current research job is being extracted. But often, the source will be either lost or stored in an individual’s bookmarks. Others in the team who could build upon the identified source don’t have access to it.

A real information management challenge. Typically this challenge has two answers, the first one would be to ask every researcher to qualify any newly discovered useful resource – this would include adding it to one or more predefined categories, writing an abstract of the content, rating it according to different metrics and the like. The second, and much more feasible answer, would be to use the Slense platform, simply submit any content to it that seems relevant – without the need to further describe the content – Slense will visit the page, index it, extract relevant terms and provide it to any colleague that searches for a fitting keyword. Over time the team build up its index of domain specific and relevant sources and can regularly build upon what the team has already explored in the past.

One more way forward for Slense – we are currently working on a special solution for companies and teams with their specific requirements in mind. Stay tuned for more on this!

According to Forrester Research, which published the report about the Social Web and a strategy to analyze the market’s social technology behaviour in 2007 , the participation behaviour on the social web can be segmented into at least six categories. Each category represents one rung, beginning with the inactives, the spectators, who are just consuming the web with nearly no participation, the joiners and collectors, who are active in social networks, the critics, who are commenting and reviewing and the creators, who are publishing content on the web. In 2007, the vast majority could be found on the lowest rung (inactives – 52%), followed by the spectators (33%).

At the beginning of 2010, Forrester Research released an update to the report, reflecting the development and popularization of social networks and -media. While in 2007, the vast majority of consumers were inactive (i.e. not participating in social networks of all kind), in 2010 the numbers grew on all rungs above the inactives. Accordings to the released update, spectators gained a considerable increase of 37%, while inactives lost about 35%. Besides updating the statistical data, they also introduced a new rung just below the creators, the “Conversationalists”. Conversationalists are people, who are regularly updating their status on social networks, wich are proividing such a feature (like Facebook or Twitter).

We’re pretty sure these numbers will shift towards the upper rungs over the next years, letting people create more value of the social networks around (e.g. using the Open Graph protocol – see below, etc.).

We’re looking forward to seeing new developments with keen anticipation and preparing Slense to support you climbing the social ladder!

This instantly reminds me of what David Weinberger called “lumps and splits” in his book Everything is Miscellaneous. The key point is that in our contemporary media environment, knowledge became something very subjective and individual. Knowledge can be categorized (i.e. lumped and split) arbitrarily by different groups of people depending on their needs and interests. Take Italian operas for example. A layman music enthusiast employs totally different categories, concepts and terms than a professional music researcher. Depending on what you’re up to, you might lump and split the vast information that’s out on the web in different ways.

We’re excited that we’re seeing this new way of interacting with knowledge in our data. The next step for us is to allow users to create different profiles all within a single account of course.