|
ÀÁÐÀÊÀÄÀÁÐÀ (Òîæå ñàìîå íî â ÷èòàåìîì âèäå)
Ècnjhbxtcrb ckj;bkjcm, xnj gjbcrjdst cbcntvs bcgjkmpjdfkb eghjotyyst
vjltkb lkz bpdktxtybz cbuyfkjd lkz hfy;bhjdfybz b fynbcgfvf. Ïj vtht
hjcnf nhfabrf b rkbrcnhbvf cnfk djpvj;ysv gtht[jl r ,jktt htfkbcnbxysv
vjltkzv. Ífghbvth, yf cvtye vjltkb ckexfqyjuj ,ke;lfybz PageRank ghbikb
vjltkb extnf gjdtltybz htfkmys[ gjkmpjdfntktq (Browserank b fyfkjubxyst fkujhbnvs).
Êjytxyj, 'njn gtht[jl yt jpyfxftn ,tpjujdjhjxyjuj jnrfpf jn
nhflbwbjyyjuj PageRank, yj jpyfxftn evtymitybt tuj drkflf d hfcxtn
htktdfynyjcnb ljrevtynf d gjkmpe yjds[ djpvj;yjcntq.
Âf;yj, xnj htfkbcnbxyst vjltkb j,tcgtxbdf.n yt njkmrj kexibq cbuyfk d
hfy;bhjdfybb, yj b gjpdjkz.n 'aatrnbdyj gjlfdkznm cgfv. Ðfccvjnhbv
ytrjnjhst gjl[jls, jge,kbrjdfyyst d cnfnmt «Identifying Web Spam with User Behavior Analysis», Tsinghua University, Beijing, 2008.
Àdnjhs htibkb ldt pflfxb:
1. Âszdktys gjdtltyxtcrbt if,kjys, gjpdjkz.obt 'aatrnbdyj j,yfhe;bdfnm cgfv,
2. Ñjplfyf gkfnajhvf lkz j,yfhe;tybz yjds[ cgjcj,jd cgfvf.
Òt[ybxtcrjq ,fpjq lkz 'rcgthbvtynf gjcke;bk ahfuvtyn kjuf gjbcrjdjq cbcntvs sogou.com
pf 57 cenjr (ktnj 2007 ujlf). Ýnjn vfccbd lfyys[ cjlth;fk 22.1 vbkkbjyf
gjkmpjdfntkmcrb[ ctccbq b 2,74 vbkkbfhlf rkbrjd gj 800 vbkkbjyfv
ljrevtynjd.
Øf,kjys, [jhjij [fhfrnthbpe.obt cgfv
Äjkz seo-nhfabrf yf ljrevtyn
Îghtltkbv ljk. seo-nhfabrf (search engine oriented visit, SEOV):
Ãbgjntpf ghjcnf: yf cgfvyst ljrevtyns gjkmpjdfntkb j,sxyj gjgflf.n
njkmrj xthtp gjbcrjde. dslfxe. Ífghjnbd, yf rfxtcndtyyst ljrevtyns
j,sxyj ceotcndetn yt seo-nhfabr. Ïhtlgjkfuftv, SEOV lkz cgfvys[
ljrevtynjd ,eltn ,jktt dscjrbv. Ïjcvjnhbv yf hfcghtltktybt rfxtcndtyys[ b
cgfvys[ ljrevtynjd gj bynthdfkfv SEOV:
Âblyj, xnj 82% [jhjib[ ljrevtynjd gjkexbkb vtytt 10% nhfabrf bp
gjbcrjds[ cbcntv. Ñ lheujq cnjhjys, lkz gjxnb 60% cgfvys[ ljrevtynjd
ljkz seo-nhfabrf 40% b ,jktt. Ïhb 'njv dctuj 1% rfxtcndtyys[ ljrevtynjd
bvttn SEOV ,jktt 70%.
Äjrevtyn rfr bcnjxybr nhfabrf
Ïhb rkbrt gj ccskrt b bcnjxybr, b wtktdfz cnhfybwf gtht[jlf abrcbhe.ncz
d web access log’t. Ë.,jq ljrevtyn vj;tn zdkznmcz rfr gjkexfntktv, nfr b
bcnjxybrjv nhfabrf. Õjnz cgfvyst ljrevtyns vjuen cjlth;fnm ,jkmijt
rjkbxtcndj bc[jlzob[ ccskjr, jyb j,sxyj yt gjhj;lf.n nhfabrf yf wtktdst
cnhfybws.
Îghtltkbv ljk. ckexftd, d rjnjhs[ ljrevtyn zdkztncz bcnjxybrjv nhfabrf (source page rate, SP):
Èp hfcghtltktybz ljrevtynjd gj ghbdtltyyjve rhbnthb. dblyj, xnj SP lkz rfxtcndtyys[ cnhfybw j,sxyj ,jkmit, xtv lkz cgfvys[:
Ïjxnb gjkjdbyf cgfvys[ ljrevtynjd, ghbcencnde.ob[ d training set’t,
htlrj dscnegf.n bcnjxybrjv nhfabrf (SP < 5%). Ëbim 7.7% cgfvys[
ljrevtynjd ltvjycnhbhe.n SP ,jktt 40%, ljkz rfxtcndtyys[ ljrevtynjd d
'njv ;t lbfgfpjyt SP — ,jktt 53%.
Äjkz rjhjnrb[ dbpbnjd
Îxtdblyj, rjyntyn cgfvys[ ljrevtynjd yt cnbvekbhetn gjkmpjdfntktq
ghjdjlbnm vyjuj dhtvtyb yf cfqnt. Îghtltkbv ljk. rjhjnrb[ dbpbnjd
(short-time navigation rate, SN rate):
Ïthtvtyyfz N vj;tn dfhmbhjdfnmcz, bccktljdfntkb ecnfyjdbkb tt hfdyjq 3.
Ôbpbxtcrbq cvsck SN ghjcn — 'nj ljkz ctccbq, d rjnjhs[ ,skj ghjcvjnhtyj
vtytt N ljrevtynjd cfqnf.
Âblyj, xnj ljkz rjhjnrb[ dbpbnjd gjpdjkztn ytgkj[j htibnm pflfxe dszdktybz cgfvf:
Àkujhbnv j,yfhe;tybz cgfvf, jcyjdfyysq yf fyfkbpt jcj,tyyjcntq gjdtltybz gjkmpjdfntktq
Âszdktybt cgfvf — nbgbxyfz pflfxf rkfccbabrfwbb. Èccktljdfntkb
bcgjkmpjdfkb yfbdysq ,fqtcjdcrbq rkfccbabrfnjh b hfccvjnhtkb jlyj- b
vyjujafrnjhye. vjltkb. Ènjujdfz aeyrwbz jwtyrb dthjznyjcnb ljrevtynf
,snm cgfvysv:
Ätnfkb htfkbpfwbb ljcnegys d bc[jlyjq cnfnmt.
Èynthtcyj, xnj ghtlkj;tyyst afrnjhs jrfpfkbcm ghfrnbxtcrb ytpfdbcbvs:
Ïj-dblbvjve, 'nj cdzpfyj c hfpkbxyjq ghbhjljq bcnjxybrjd lfyys[.
Àkujhbnv dszdktybz cgfvf:
1. Ñ,jh kjuf,
2. Ðfcxtn SEOV b SP lkz rf;ljuj ljrevtynf,
3. Ðfcxtn SEOV b SP lkz rf;ljuj cfqnf (echtlyzz ljrevtynyst lfyyst g.2),
4. Ðfcxtn SN lkz rf;ljuj cfqnf,
5. Ðfcxtn dthjznyjcnb ljrevtynf jrfpfnmcz cgfvysv.
Ðtpekmnfns
Î,exbd rkfccbabrfnjh, hfphf,jnxbrb fkujhbnvf ghjntcnbhjdfkb tuj yf
ckexfqyjq ds,jhrt bp 1564 cfqnjd. Àctccjhs cjxkb 345 cfqnjd cgfvysvb,
1060 yt cgfvysvb, 159 — pfnhelybkbcm jwtybnm. Ïjcnhjtyyfz ROC
bkk.cnhbhetn, xnj SP b SEOV gjpdjkz.n 'aatrnbdytt j,yfhe;bnm cgfv, xtv
SN:
Èynthtcyf ghj,ktvf crjhjcnb htfrwbb yf gjzdktybt cgfvf. Òhflbwbjyyj yf
dszdktybt cgfvf nht,etncz dhtvz. Ýnj [jhjij dblyj yf cktle.otq rhbdjq:
Ïhtlkj;tyysq fdnjhfvb vtnjl gjpdjkztn ecrjhbnm j,yfhe;tybt cgfvf.
Ïhfrnbxtcrbt htrjvtylfwbb
×nj,s cybpbnm dthjznyjcnm hfpvtnrb cfqnf rfr cgfvyjuj, ye;yj:
1. Äevfnm j cxfcnmt gjkmpjdfntkz:
- Ðfpvtofnm gjktpysq rjyntyn b cthdbcs
- Ñcskfnmcz yf fdnjhbntnyst bcnjxybrb
- Î,tcgtxbdfnm elj,ye. yfdbufwb.
2. Ñnhtvbnmcz gjkexfnm nhfabr bp hfpkbxys[ bcnjxybrjd
3. Ít ghbdktrfnm gkj[j rjydthnbhe.obqcz nhfabr:
- c ybprjrfxtcndtyys[ b/bkb ytntvfnbxtcrb[ htcehcjd
- gj j,]zdktybzv bkb ccskrfv, yt htktdfynysv frwtgnjhe
Ít bcgjkmpeqnt cgfv, ghbdktrfqnt wtktde. felbnjhb., wtybnt dhtvz dfib[ gjkmpjdfntktq. Ólfxb! seonews
|
|