5-1393msg2.txt

该程序实现的是一个垃圾邮件过滤系统

Subject: re : 5 . 1370 can the comparative method be apply to syntax ?

theoretically , there be no reason why the comparative method could not be apply to syntactic property . however , there be two practical , very great , impediment : 1 . syntax deal with word order , what be word order ? and further which word order in which environment ? e . g . : ich verstehe sie . ploetzlich verstehe ich sie . ( subordinate conjunction ) ich sie verstehe . un grand homme . un homme grand . ma chemise propre . ma propre chemise . 2 . how do we measure the difference ? how far be svo from sov ? from vso ? not only we do not know how to describe syntactic property satisfactorily , but we do not know how to measure their difference , or , it boil down to the same , their similarity . now , you can validly argue that a binary scale ( same , different , and no shades of grey in between ) be good enough . remain that you do not reach valid comparison on the basis of a single feature , but of many . if language x and y have n identical feature out of n , we be tempt to say that their syntactic similarity be n / n . but that be true if and only if each feature have the same weight . on top of that , we do not even know what constitute * one * feature . for instance , we may consider svo as consist of two to three feature : 1 . relative order of subject and verb 2 . relative order of verb and object 3 . relative order of subject and object where necessary to disambiguate . and further , how do we specify free sv order for instance ? via a ternary relation ( precede , follow , free ) , thus order = ( free , s , v ) or a binary relation ( precede , follow ) and a union , thus order = ( precede , s , v ) + ( follow , s , v ) ? too many question unanswer , i fear even unaddress . too many question issue from them . we cannot , i think , find a proper methodology on such shaky grounds . we 'd better be content with argue futilely until we be blue in the keyboard . or else set about tackle those question , a daunt task . j . guy @ trl . oz . au