An Unsupervised Approach based on Fingerprinting to the Web People Search task

In this paper we present the results obtained in the Web People Search task (WePS-2) when using an unsupervised approach based on document fingerprinting techniques. In the context of document indexing/retrieval, we consider a document fingerprint to be a specific code which may be used to uniquely identify this document from the rest of the text collection. In terms of implementation, we consider a document fingerprint may be obtained through hash-based indexing. The evaluation of the experiments carried out show that the implemented technique could have a positive impact in the analysis/indexing of huge volumes of information. However, the feature set for all the documents in the WePS framework needs to be further investigated.