A family of link based metrics for the evaluation of Web documents

Abstract

1 1ntr0duct10n 7he exp10510n 0f the we6 ha5 6een a maj0r c0ncern am0n9 we6ma5ter5 and c0n5e4uent1y, eva1uat10n 0f hyperd0cument5 ha5 6ec0me a maj0r 155ue. 7h15 ha5 6een pu5h1n9 the deve1-0pment 0f many d1fferent type5 0f t0015 a1m1n9 at 1mpr0v1n9 the 4ua11ty 0f hypermed1a de519n. 1t we11 kn0wn that we6 51te5 de519n 15 n0t an exact pr0ce55; a5 the de519n act1v1ty Pr09re55e5, many d1fferent a5pect5 c0ncern1n9 the 501ut10n 0f the pr061em 0f pr0duc1n9 a • • 900d" hyperd0cument 5h0u1d 6e c0n51dered. 51nce hypermed1a de519n pr0v1de5 many d1fferent and a1ternat1ve way5 0f 0r9an121n9 1nf0rma-t10n, there are n0t 5tr1ct ru1e5 wh1ch 5h0u1d 6e f0110wed 1n 0rder t0 acc0mp115h a • • 900d" de519n. 7he ex15tence 0f pract1-ca11y n0 ru1e5 a60ut what 5h0u1d 6e taken 1nt0 acc0unt when creat1n9 a hyperd0cument make5 th15 ta5k hard and the 1ack 0f c0n5en5u5 0n h0w t0 ap-pr0ach 1t make5 the ta5k even harder. May6e th15 15 0ne 0f the rea50n5 f0r the ex15tence 0f 50 many p00r1y wr1tten hyperd0cument5, 1ack1n9 an we11-0r9a-n12ed 5tructure. 50me pr0pert1e5 0f hyperd0cument5 that can 6e u5ed f0r 1dent1fy1n9 p00r1y de519ned hypertext5 have 6een p01nted 0ut 1n [7h15t1ewa1te-95]: un50und 11nk5, n0 c0n515tent cr1ter1a f0r 11nk creat10n, 1nc0mp1etene55 and d1ff1cu1t t0 ma1nta1n. 1t 15 rec09n12ed that ma1ntenance 0f we6 51te5 15 a650r61n9 a 9r0w1n9 p0rt10n 0f the 50ftware 6ud9et and that 1t 15 n0t a1way5 pr0per1y p1anned 6y 50ftware deve10per5. 0ne 0f the rea50n5 f0r that 15 the fact that the WWW 9ather5 a hu9e am0unt 0f 1nf0rmat10n 1n a very dynam1c, f1ex161e and d15tr16uted env1r0nment; 1t 15 c0n5tant1y under chan9e a1m1n9 n0t 0n1y at 1t5 expan510n 6ut a150 at the c0n515-tency 0f 1t5 c0ntent5. 1n fact, the pr0-ce55e5 0f creat10n and ma1ntenance 0f hyperd0cument are re1ated t0 each 0ther due t0 the f1ex161e nature 0f 11nk5; they a110w 1nf0rmat10n t0 6e ea511y re0r9a-n12ed. 7he c0mment 5tated 1n [E6ert-96] c0n-cern1n9 50ftware 1n 9enera1, • • 60a15 5uch a5 4ua11ty 1mpr0vement 1n term5 0f 1ncrea5ed re11a6111ty and ma1nta1na6111ty are 0f 9r0w1n9 1mp0rtance 1n 50ftware deve10pment" can a150 6e app11ed t0 hypermed1a auth0r1n9 pr0ce55. 7he5e 90a15 can 6e ach1eved 6y u51n9, dur1n9 hypermed1a auth0r1n9 pr0ce55, a t001 that pr0v1de5 feed6ack 0n the hyperd0cument 5tructure. We 6e11eve that the pr0per u5e 0f 11nk 6a5ed metr1c5 wh05e va1ue5 c0u1d 6e per10d1ca11y c011ected 6y an aut0mat1c t001 can c0ntr16ute t0 a 5ucce55fu1 man-a9ement 0f hyperd0cument auth0r1n9 pr0ce55 and we6 ma1ntenance. L1nk 6a5ed metr1c5 can 6e u5ed f0r 1dent1fy1n9 50me …