Entity　recognition　is　the　basis　of　text　mining.　With　the　further　development　of　knowledge-driven　applications,　types　of　target　entities　are　increasingly　subdivided.　The　lack　of　corpus　and　the　limited　number　of　entity　have　been　the　main　challenges　of　entity　recognition.　Based　on　this　observation,　this　paper　proposes　a　weak-supervision　method　for　recognizing　entities　from　a　specifically　narrow　domain　by　fusing　domain　relevance　measurement　and　context　information.　The　experimental　result　shows　that　the　proposed　method　has　high　efficiency　and　accuracy　without　manual　participation.