The　reading　comprehension　(RC)　task-accepting　arbitrary　text　input　(a　story)　and　answering　questions　about　it.　The　RC　system　needs　to　draw　upon　external　knowledge　sources　to　achieve　deep　analysis　of　passage　sentences　for　answer　sentence　extraction.　This　paper　proposes　an　approach　towards　RC　that　attempts　to　utilize　semantic　information　to　improve　performance　beyond　the　baseline　set　by　the　bag-of-words　(BOW)　approach.　Our　approach　emphasizes　matching　of　linguistic　features　(i.e.　verbs,　named　entities　and　base　noun　phrases)　and　semantic　extending　for　answer　sentence　extraction.　The　approach　gave　improved　RC　performance　in　the　Remedia　corpus,　attaining　HumSent　accuracies　of　41.3%.　In　particular,　performance　analysis　shows　that　a　relative　performance　of　19.7%　is　due　to　the　application　of　linguistic　feature　matching　and　a　further　10.3%　is　due　to　the　semantic　extending.