Lexical Effects in Verb Sense Disambiguation.pdfVIP

F0;lB?5A@-2rC$K$*$1$k8lWC8rBX8=]Timothy Baldwin, FA1J7r?-, EDCfJf@Q El5~9)6HBg3XBg3X1!pJsM}9)3X85f2J 35MW K\85 f$G$O!V8lWCBN7O7k9g2A-=q!W$J$I$N$h$$JA*Br@)Ls%Y!%9$N-=q$rMQ$$$?F0;lB?5 A@-2rC%7%9%F%‘$rDs0F$9 $k!%K\jK!$G$O!$3F!9$N3JMWAG$KBP$7!$A*Br@)Ls$,$I$NDxEY87L)$KF/$/$N$+$H$$$C$?EY9g$r%7%=!%i%9$NLZ9=B$$K $h$jI2A$9$k!%$3$l$K$h$j!$3JMWAG$rJN,$5$l$?;vNc$KBP$7$F$b8l5 A3N?.EY$r8zN(NI$/7WNL$G$-!$8l=g!I=AX3J8r BX$K$bMF0W$KBP1~$G$-$k!%V8+$k!W$rCf?4$H$9$k148 8D$NF0;l-=q$rMQ$$!$F1F0;l$r4 $?‘291 8D$N@a$rB?5 A@-2rC$7 $?$H$3$m!$92.10% $N@52rN($,F@$i$l$?!%Lexical Eects in Verb Sense DisambiguationTimothy Baldwin, Takenobu Tokunaga, Hozumi TanakaTokyo Institute of Technologyftim,take,tanakag@cs.titech.ac.jpAbstractFirst, we propose a uni ed framework for evaluating verb sense in a selectional restriction-based dictionaryarchitecture, including both generalised and xed verb senses. The proposed methodology is primarily basedaround the restrictiveness of the selectional restrictions on each case slot, evaluation of which is made throughcalculation of the average depth of semantic classes governing the case slot in question. As a spin-o of thisrestrictiveness measure, we develop the ability to predict likelihood of sense in the case of underspeci cation,and go on to combine the measure with methods to variously handle word order and surface case variation. Byway of experimentation, a test dictionary of 148 verb entries lexically deriving from the Japanese verb mi(-ru)\(to) see, is used to empirically evaluate the eectiveness of the overall system on an open test set of 291clauses. The overall accuracy on this test set is 92.10%.1 IntroductionWord sense disambiguation (WSD) refers to the taskof identifying the sense of a term within a given lexicalcontext, from within a set of sense candidates. Ourparticular interest in WSD lies in verb sense disam-biguation (VSD), which in the case of Japanese, is di-rectly equivalent to full WSD in terms of complexity,as nite-in ected verbs can