This project sets out to achieve two goals. The first objective is to update the annotation system for Red Hen’s NewsScape dataset to FrameNet 1.7 using Open-Sesame and Semafor parsers. The second objective is to expand the lexical units, frames and frame-to-frame relations in FrameNet 1.7 through a knowledge-driven approach and a distributional semantics approach. The knowledge-driven approach uses BabelNet to induce the frames of unrecognized lexical units in the tagged NewsScape dataset. The latter distributional semantics approach uses Deep Structured Semantic Models (DSSM) to create word embeddings of lexical units (LUs) to resolve the inconsistency in FrameNet hierarchy, tag LUs with their missing frames, and locate new frames using SemCor corpus. If time permits, DSSM is used to expand the frame-to-frame relations with Entity and Event frames using ACE 2005 Entities and Events dataset.