Combination of an Automatic and an Interactive Disambiguation Method

In llatllral language processing, many methods have been proposed to solve the ambiguity l)roblems. In this paper, we propose a lechnique to COlnbille a method of interactive disamlfiguation and automatic one tbr anlbiguous words. The characteristic of our method is that the accuracy of the interactive dis-alnbiguation is considered. The method solves the two following l)roblems when combining those dis-amlfiguation nwthods: (1) when should the interact .iv(" disaml)iguation be executed? (2) which ambiguous word shouhl be disambiguated when more lhan one aml)iguous words exist in a sentence? Our method defines the condition of executing the interaction with users and the order of disambiguation based on the strategy where the accuracy of the result is nlaximize(I, considering the accuracy of 1.11o intel'acl.iw" disalnbiguation and autonmlic oil(', (ls-ing this me{hod, t,ser interaction can be controlled while holding the accuracy of results. 1 Introduction In natural language processing, many methods hay(" been proposed to solw~ the alnbiguity prob-lems(Nagao and Maruyama, 1992). One of those technique uses interactions with users, because it. is dilticult to make all the knowledge for disambigua-lion beforehand. That technique is classitied into two l.ypes according to i.he conditkm of executing user interaction. One type(TypeA) is that the dis-anfl)iguation system executes interactions{ Hlanchon et al. In this paper, Ty-l)eA will be adopted 1.inca.use Typel3 gives users more trouble than TypeA does. For example, in TypeB, a user may have to find where is wrongly analyzed in inl)ut sentences. In TypeA, the two following conditions must be determined: (1) when should interactive disam-biguation be executed? (2) which aml.)iguous words should be disalnbigual.ed when more than one a.lll-biguous word exist in a. sentence? Considering the accuracy of the analyzed result, they should be decided by both the accuracy of the interactive dis-ambiguation and that of the autonlatic ¢lisambigua-tion. The traditional methods did not considered the accuracy of the interactive disambiguation. For instance, the accuracy of the analyzed result may decrease in spite of execuling lhe user interaction if the accuracy of the interactive disalnbiguation is lOW. In this l)al)er, we l)ropose lhe method to combine the intera('liw~ disalnbigualion and the aulo-marie one, considering each accuracy. The melhod allows the disambiguation system to maximize Iho accuracy of the analyzed resull. This paper focuses on the ambiguity caused by ambiguous words that [l~-IVO lllOr(" t[l~:lll ()lie lllealliltg. Section 2 rol)resell|S precondilions for disambiguation. In Section 3, we describe 1,he condition of executing the inleracl.ive …