The Evolution Of Machine-Tractable Dictionaries

I, hltli~ttut:lion T i l e u s e l u l n e s s o f Mach ine -T rac tab le Dict ionar ies (MTDs) ill atlCtilllatic SCllSe taggillg el running text ix clearly demons t ra ted in Tong , el ill., (1993). While previous work (Yarowsky, 1992: Gale, el al., 1092, 1903) lelies heavi ly on tile role of statistics, TOllg'S sysienl illilkes use of (Ill(: MTD as well as m<~ Machine Readable l)iciionarles (MRI)s) in gin examplebased reasonillg process ill laggint~, sellsC Ilillllbers l() l lovel words, eompol lnds ~,\,ords, gilld pJllHSes in tile input text. The hit rate of correcl ,~,Cll:-;c lagging rtlnS ax high as 90%. The lewes! IlJl rtll¢ c:tcr iecordcd was 70%. The sysieill is eotlsidercd a necessaiy inechanisin for the construc ihm ()1 annotated Chinese ]~#ollilor C.'orlJora (SillCILIIC, If) t) I) I1o111 IlllllliIllL (;hilleSe tc',t . l'~girlier work (Wilks, et al., 1'090) identified a dislii101ion belwcel l MRI )s and NqTJ)s. Th is eal×'r recogll izes a further dist inction between two types el MT1)s> i.e., Type I and Type il MTDs. Type l MTI)s arc bui l t ()ll the def in i t ion pr i in i i i vcs (fi on0 pgiri i0ulal M R D wtlerea,,-; Type I1 MTI)s are btl i l t ',In the semanli0, prmli t ives of one |lgiiciellJar ilgiCtlr:.ll IgillpIlage. The MI 'D funct i tnl ing il l T,,mg's sysCenl belongs co die l ' ype I category. The evolut ion fr lmi Type I t(~ Type I1 MTI)s iopiesents all evoitiCiOli ()1 the i)r i init ives ti~c MT1)s arc constrtlcted o11. Th0 atithors bel ieve that, although it is leasible lo derive gi l lall lral sol (11 de l in i l ion pri lnil ixc~ fr(lill o11o parlicti[gir MRD, a descriptive set of Selllantl0 primitives (Wiiks, 1977) of Olle par i icuh i r imlura] ILingilat2e is prefcigibly deri~ed irl)i l l inol'e than t)llC ila[tll'aJ s(ltlrc0 Sllch :t.s :1 diction',u'y. T h i s p.os;ilip.n rcpresellis gi bgickoll frol l l the previous olle Ihgil ctailns tile dcr iva l ion of a natural set el senlanli0 piinlici ',es lrol l l oile part iet l iar MRI), par th :u lg i r i ) lhe Jz)#1~,l#1fHl I)i¢lionarv (!f ('olllettl])orarv l','ngli~h (I ~l)O(:t",) ((]uo. 1989). Three Type l M T i ) s 0olnplcJcd al TsiTI!.',hli~ (] l l i~,clsity w i l l bc presented. 'l hose include MINIIA)OCI); consii-tie/ed for I,oilgingin l ) ic i ioanries, lhc M T i ) versiou of 7he Mo&'r. (?hillese Di¢'liOllflr)~ ¢!/" Gem, ra l ('[lill~'A'e ( ' ] l t lraf ' l t ' rs ( M'['I)X I A N'I ' ()NG) (1 ;il, It)leT)> and Ihe M T I ) VClSi(lll oJ" 'lTle Ml l / l i Jhm' l iona l I)icliollarv o/ Moderll (.'hillese IVord~ (MTD I ) [J()(K)N(;NI",N(~) (l:eng & Zh(lu, I t)~O). l i f l m t s to dcri~e natural ~;0ts ~)1 semant ic pri lni l ivos l l l /nl Typt; I MTDs girt Ihcn des0rihcd These inclll(le work (Ill detccting 0Jr0 tl]ar dcl ill tiOllS wit ln n t)llC parliciilgir MT1), work on colnpl l lh lg the coinpai i l~i l i iy I/I word SellSCS beCw0cfi IW() IIW~noJillgtlal di0tiongirics, and the most recent step at der{vmg a ilatura] set o! Chincse senlaniic pr imi l ives Into1 Typ0 i M'I 'Ds (11 (Jhinese. The work ~as done on ~tii1.4 worksigii ions with the (7 eoinF, tlter lallgUgige.