Air　pollution　has　become　a　substantial　environmental　issue　in　China,　which　has　seriously　affected　human　health.　To　clarify　the　relationship　among　air　pollutants,　pollution　sources,　influencing　factors,　evaluation　indicators　and　harms,　it　is　essential　to　build　a　domain　ontology　for　air　pollution.　Domain　ontologies　have　been　gradually　accepted　as　a　method　to　indicate　the　relationship　between　terms.　However,　there　is　not　a　complete　ontology　in　air　pollution　domain,　and　building　domain　ontologies　manually　is　time-consuming　and　inconvenient.　In　this　paper,　a　semi-automatic　approach　is　presented　to　build　an　ontology　in　air　pollution　domain.　This　paper　proposes　a　method　of　entity　relationship　joint　extraction　based　on　attention　mechanism,　which　is　combined　with　core　concept　mining　method　to　extract　knowledge,　and　then　concepts,　relationships,　and　relevant　instances　are　organized　in　hierarchy.　Finally,　the　ontology　model　is　constructed　semi-automatically　and　semantic　inference　is　also　carried　out.　The　research　shows　that　this　knowledge　extraction　method　can　avoid　the　error　accumulation　caused　by　entity　recognition　and　relationship　extraction　and　deepen　the　inner　connection　between　them,　and　a　large　amount　of　effective　knowledge　in　the　field　of　air　pollution　can　be　extracted.　In　addition,　the　ontology　constructed　based　on　this　method　can　also　visually　analyze　the　relationship　between　various　classes　and　concepts　in　the　field　of　air　pollution,　deduce　the　pollutant　propagation　path,　and　provide　practical　experience　for　the　semi-automatic　ontology　construction　in　other　fields　and　data　support　for　further　research.　©　2020,　Springer　Nature　Singapore　Pte　Ltd.