汉语言网络统计特性与半监督文档聚类算法分析.pdfVIP

A b str a e t W ith th e exP onentially inereasin g of eleetronie do eum ents on the we b ，there ex- ists u rg ent n eed to P ro eess th ese d o eu m en ts a u to m atiea lly u sin g eo m P u ter ，in e lu d in g au to m a tie e la ssif iea tio n ，e lu sterin g a n d su m m a rizatio n o f d o eu m en ts. In th is P ap er ，w e foeu s o n d o eu m e 毗 e lu s tering . D o eu m e ltl elu s terin g eo n s ists o f th ree P ar ts : d o eu m ent reP resent 就 io n ，elu sterin g a lg o rith m a n d P erform a nc e eva lu a tio n ，o f w h leh d o eu m ent reP resen tatio n an d clu ster in g a lg o rith m are th e m o st im P o rta nt steP s. Ve etor Space M odel(V SM ) dom inates in the doeum ent representation . D ue to the sem i一stru etu re ch aracteristie o f d o eum ents，it 15 easy to see the d raw b acb o f V S M . R ecent l乳 eo m P lex ne tw o rk ，a s a to o l for stu d y in g eo m P lxe sy s tem ，h a s attrac ts a fot o f attent io n fro m res eareh ers. D o eum- ent s，a s th e wr itten 一lan gu ag e o f h u ln an b e in g ，a lso ean b e regarded as eom P lex sy stem . 节几 stu衍 th e statistiealp roP ertles of th e C h ln ese lan gu ag e n etw o rk ，w i th in th e fra m ew o rk o f eo m P lex ne t酬叼rk ，in o rd er to sh ed som e n ew sig ht s ab o ut d o eum ent reP resenta tio n . B ased o n o n e o f th e larg est C h in ese eo rP o ra ，i.e .， P eop le’s D aily C orpuS ，we ~ truct t喇勺netw orks (C LN I an d C LN Z)from tw o different res P ec ts ，w i th C h in ese wo rds as no d es. In C L N I ，a lin k b etw een two n o d es e廊 ts ifth ey aP P ea r n ex t to each o th e r in o n e se ltlen ee ; in C L N Z ，a link reP resent s th at tw o n o d es ap P e ar s