Many event models are available for event extraction. These models have event information for agent, time and place. Effort to collect information for similar events to construct a lexical database is yet to be carried out. This research intends to achieve such objective by proposing an automated event extraction approach based on semantic role label (SRL) predicate argument structure (PAS). Events which are indicated by predicates with similar sense are gathered to collect event information for lexical database. This is a generic approach which does not need any event model. This paper has described and illustrated how SRL PAS is mapped to event information to create the lexical database.