一种基于字同现频率的汉语文本主题抽取方法_马颖华.pdfVIP

4 0 6 Vol14 0, No16 200 3 6 JOU R NA L OF COM PUT ER R ESEA RCH AN D DEVEL OPM ENT J une 200 3 马颖华 王永成 苏贵洋 张宇萌 ( 200030) ( m a- yinghua@ sjtu1 edu1 cn) , 1 1 , , 1 , 1 , , , , 1 , 77119%1 , 1 T P39111 A Novel Chinese Text Subject Extraction Method Based on Character Co-occur- rence M A Y ing- Hua, WA NG Y ong- Cheng, SU Gu-i Yang , and ZHA N G Yu- Meng ( Dep artment of Comp uter Science and Engineering, Shanghai Jiaotong University , Shanghai 2000 30) Abstract Subject ext ract ion is one of t he fundament al w orks of nat ural language processing 1 Word seg- mentation or word ex tract ion is alw ays the first st ep of subject ext ract ion1 As t here is no int ervals among w ords in Chinese t ex t, both w ord segment at ion and w ord ex tract ion are dif ficult 1 In this pap er, a novel Chinese t ex t subject ex traction met hod based on ch aracter co- occurrence is put forw ard1 N eit her word seg- mentation nor word ext raction is required in t his method1 T he method has high processing speed and can be used in bot h subject sentence ex tract ion and subject p arag raph ex tract ion1 A not her advant age of t his ap- proach is t h at it can be used to process not only Chinese text but also t ex t in ot her languages and even mult-i languag e text 1 Results of experiment s show t h at the approach gain s high accuracy of 77119% in mult-i st yle t ext of new s1 A nd w it hout w ord seg mentation, t he accuracy does not decline1 Key words nat ural langu age processing subject ex tract ion co- occurrence , 1 1 ( [ 1, 2] [ 3]