ThinkingNectar

Spend the good part of Saturday at Barcamp 3 held in Ngee Ann Polytechnic.

Was there fairly early just before any talk started. Saw usual faces and of course some new faces as wel.

First thing in the morning was a presentation by Bernard Leong, covering for the first time, in his own words, a non-venture funding presentaton on some technologocial ideas which he feels will have an impact.

Bernard presentating at Barcamp singapore 3

2 items will be covered and the first being a weighted connector implementation of social network. Currently most social network are implemented via a binary concept. A friend (a connection) or not a friend (no connection).  Such a connection does not measure or captured the value of the relationship. Be it a siblings, friends only, mentor/student.

The concept presentated in in a simple form. An implementation of 1 to 10 was presentated. Such a weighted and directed (some computing jargon but do bear with me) will offer insight to the strength of a particular connection between 2 entities in a social network. And such insight can helps people in both personal and business to make introduction for a connection. Getting a introduction or recommendation from a close friend will yield more mileage than from someone who is just an acquaintance.

Bernard added that such a network should offer some form of calibration to prevent gaming.

Having such a weighted network will bear more value than the current social network. To add to the concept, a major concern to consider is the transitional state of relationship. For example I may rate a college friend highly with a 9, but after graduation, the value should “decay” as we do not keep in touch and are generally busy with our professional life. That is the general scenario. Then for a more drastic scenario when a business partnership suddenly turn south, such weight in the network graph may not be updated and will offer misleading dimensional data.

The perils in modelling imperfect relationship between humans.

The second item that was cover was voice recognition on mobile devices, primarily handphone. Mobile devices with it lesser computing power will require a more elegant solution. Bernard proposal was a open source project for such a particular need. In addition, the solution will also include localisation of local accent of different people across different countries. He mentioned too that localised accent is a major problem facing voice recognition at the moment.

In addition, the solution will also individual to upload their voice profile to a centralise database to form a global pool of voice data which can then be processed (by who ever needs it) for accent profiling and such. The central database can also be alternatively used as the OpenID for voice.

Next talk which I attended was one of Social Media Marking for business which personally sound too guruish and I do not buy much into it. Guess I should be avoiding how-to talks in Barcamp.