In this paper we describe our participation in Web People
Search Clustering task. We present a new methodology
based on language models to improve Web People Search
disambiguation. In particular we introduce two different approaches:
One of them uses alternative weighting functions
to represent a document and it apply a classical clustering
algorithm. The second approach uses two sources of occupational
information as reference collections and it applies
an heuristic-based technique in order to resolve the number
of different identities. Moreover, we have studied the impact
in results of using stemming and a variant of Interpolated
Aggregate Smoothing applied to language models.