Semantic role labeling has regained importance in the past two years as a result of the availability of large annotated data from the PropBank (University of Pennsylvania and University of Rochester). Our work focuses on a similar task but uses a different set of semantic relations which unlike the PropBank are not restricted to only verbs and their arguments. This difference coupled with our generic and detailed relation set makes the use of PropBank inadequate in our case. These distinct requirements of our study necessitated the creation of our own dataset. Using this dataset we look at two feature-rich parsers as sources of syntactic features and also do statistical syntheses of semantic resources like WordNet, LDOCE and Rogets in our quest to find features which provide the best solution. We look at a hybrid solution and propose our findings as a new baseline for our version of this semantic role labeling task. The aim of this study is to look at different methods and resources that will help us progress our TANKA project from a semi-automatic to an automatic process. TANKA is a knowledge engineering aid which helps extract semantic relations from a semi-technical text as pairs of syntactic units and assigns them semantic relations.