In the post-genomic era, the organization of genes into networks has played an important role in characterizing the functions of individual genes and the interplay between them. It is also vital in understanding complex cellular processes and their dynamics. Despite advances, gene network prediction still remains a challenge. Recently, heterogeneous genomic and proteomic data were integrated to generate a functional network of yeast genes. The Gene Ontology (GO) project has integrated information from multiple data sources to annotate genes to specific biological process. Generating gene networks using GO annotations is a novel and alternative way to efficiently integrate heterogeneous data sources. In this paper, we present a novel approach to automatically generate a functional network of yeast genes using Gene Ontology (GO) annotations. An information theoretic semantic similarity (SS) was calculated between every pair of genes based on the method proposed by Resnik. This SS score was then used to predict linkages between genes, to generate a functional network. An alternative approach has been proposed using a measure called log likelihood score (LLS). The Functional networks predicted using the SS and LLS measures were compared. We discussed our experiments on generating reliable functional gene networks and concluded that the functional network generated by SS scores is comparable to or better than those obtained using LLS scores.