Natural　language　question-and-answering　is　one　of　the　most　convenient　means　for　communicating　with　the　Semantic　Web,　which　is　typically　in　the　form　of　online　knowledge　bases　encoded　in　Web　Ontology　Language　(OWL).　To　understand　a　natural　language　question,　it　is　essential　that　it　can　be　translated　into　a　query　that　is　understandable　by　the　knowledge　bases.　This　article　is　concerned　with　the　task　of　semantic　mapping　from　natural　language　questions　to　OWL　queries,　and　proposes　an　automatic　and　domain-independent　mapping　framework,　called　Three-Phases　Semantic　Mapping　(TPSM).　The　TPSM　framework　approaches　the　task　of　semantic　mapping　in　three　interrelated　phases:　(i)　formalizing　knowledge,　(ii)　building　semantic　mapping,　and　(iii)　combining　OWL　queries.　First,　formalizing　knowledge　formalizes　the　units　of　mapping　in　the　natural　language　and　OWL　knowledge.　Second,　semantic　mapping　builds　the　transverse　mapping　between　questions　and　OWL　knowledge,　as　formalized　in　the　first　phase,　by　means　of　working　with　the　so-called　Fuzzy　Constraint　Satisfaction　Problems　(FCSP).　Third,　combining　OWL　queries　obtains　valid　Resource　Description　Framework　(RDF)　models　by　applying　predefined　templates　and　their　corresponding　combining　methods.　We　have　implemented　a　prototype　semantic　mapping　system　based　on　the　framework,　and　have　conducted　a　series　of　experimental　validations　involving　OWL　knowledge　bases　in　different　domains　as　queried　by　various　types　of　questions.