Social influence ranking service PeerIndex now provides more granular rankings of users, including via the PeerIndex API. Besides the overall influence score of a user, called PeerIndex (PI), the service now provides the so called ‘topical PeerIndex’ (tPI). This tPI ranks the users by his/her influence in a specific topic domain.
PeerIndex currently distinguishes the following 8 major topics, which a user is evaluated against:
- Arts and Entertainment
- Technology and Internet
- Science and Environment
- Health and Medical
- Leisure and Lifestyle
In additional to that PeerIndex also calculates 5 niche tPI scores for the user. These are much smaller topics of which PeerIndex claims to track several thousand.
"The use cases for tPI [...] are very strong, much more so than a single overall score," PeerIndex's CEO Azeem Azhar said of this new ranking method. It is "particularly powerful for curation applications as well as personalizing web experience."
As an example of what can be done with the tPI, PeerIndex has launched PeerPerks, which it calls "rewards for being social."
This upgrade with code name “Lancelot” has also lead to an extension of the PeerIndex API, in order to offer this extended functionality to Integration partners as well. The old API Endpoint profile/profile is still available, while the new API endpoint profile/show includes the new topical PeerIndex. For illustration purposes we are listing the PeerIndex data for a request for the author’s twitter username:
The value in this new offering strongly depends on the quality of the topical classification algorithms that PeerIndex is using under the hood. "[We are using] a range of different classification code to do entity extraction and tweet classification" Azhar told ProgrammableWeb. Algorithm-wise PeerIndex uses "own code and some open source libraries." When asked further about the most technically challenging aspect of the implementation of this classifiers, Azhar mentions that PeerIndex "had to do some interesting things to handle the scale of classification, as well as use some quirky features to improve the f-score."
It is fair to assume that we will see some interesting mashups and business applications, integrating this new PeerIndex feature soon.