Weintroduceaknowledge-basedapproachto deepknowledge discoveryfrom real-world naturallanguagetexts. Datamining, datainterpretation,anddatacleaningareall incorporatedin cycles of quality-basedterminological reasoningprocesses..The methodology we proposeidentifies new knowledgeitems and assimilatesthem into a continuously updateddomain knowledgebase.