5-1286msg2.txt

运用贝叶斯网编的一个识别垃圾邮件的程序

Subject: typological method

pace fritz newmeyer , there have be extensive discussion of the methodological issue that he raise . ( these remark be independent of any assessment of johanna nichol 's book that fritz refer to , incidentally . ) first , good typological work devote a substantial amount of effort to define and delimit what sort of construction should or should not belong to a particular structural type . good example be many of dryer 's papers ( e . g . dryer 1989a , 1989b , 1992 ) , koptjevskaja - tamm 1990 : ch . 1 - 3 on nominalization , stassen 1985 : ch . 2 , 4 on comparative and ' chain ' construction , and haspelmath 1993 : ch . 2 on indefinite pronoun . more generally , the discussion of the proper typological markedness criterion in croft 1990 : ch . 4 - 6 be an attempt to provide some * general * structural criterion for a large class of cross-linguistic phenomenon ( see for example the formulation of the structural consequence of hopper & thompson 1980 's transitivity hypothesis on p . 131 , or the application of the criterion to the syntactic category problem in croft 1991 : ch . 2 - 3 ) . delimit and classify the phenomenon to be analyze be not a simple descriptive task ; in fact , it raise some of the most interest issue in typological analysis , and have hardly be ignore in the typological literature . second , with respect to basicness for a language type in particular , in the passage from croft 1990 that fritz refer to ( pp . 33-36 ) , i argue that typologist have move from classify language typologically to classify construction typologically , and that that be a good thing . nevertheless , i suggest some general criterion for determine a " basic " language type . and despite my reservation about identify " basic " language type , it have to be say that in the great majority of case , define a " language type " instead of a " construction type " be not terribly difficult . matthew dryer ( p . c . ) have observe that inspection of text generally tell you very quickly what the " word order type " of a language be . in my own experience , in a study where i be compare the syntax of prenominal and postnominal modifier , i be quite frustrate at how difficult it be to find language where a modifier freely occur both pre - and postnominally . of course , identification of " basic " or other type depend on what resource you have available to you ( quality of grammatical description , availability of text / language consultant etc . ) as well as what phenomenon you be study ( e . g . word order or head mark [ easier ] v . relative order of modifier or the syntax of nonrestrictive modification [ harder ] ) . but for most study it be not an insurmountable problem , although there alway some problematic case . third , the sample problem * have * be discuss at some length by typologist ( beside bell 1978 's seminal article , see also dryer 1989c , perkin 1989 , croft 1990 : 18-25 , rijkhoff et al . 1993 ) . the issue be too complex to go into here at length . but it should be point out that the issue fritz raise apply to different sort of sample with different purpose in mind . the question of how independent particular instance be have to do with probability sample ( see in particular the dryer , perkin and croft reference for discussion of these problem ) . note in particular that the stability of the phenomenon be study be an important factor : the less stable it be , the more likely case be historically independent . the question of come across rare type pertain to variety sample ( see in particular rijkhoff et al . ) . some study ( e . g . tomlin 1986 , koptjevskaja - tamm 1990 , haspelmath 1993 ) draw on a very large variety sample and then select a stratify probability sample from it for the appropriate generalization . dryer 's sample technique attempt to combine the need of variety and probability sample into a single sample procedure . i should also add that with the shift toward dynamic ( diachronic ) interpretation of synchronic typological pattern , the examination of cognate phenomenon in relate language , and the comparison of " nonbasic " as well as basic typological strategy , become another important method of analysis . that means that to some extent , the problem of basic v . nonbasic type in a single language , and of historical relatedness of datum set , can be avoid - - - in fact , exploit - - - in modern typological analysis . bill croft reference : bell , alan . 1978 . language sample . universal of human language , vol . 1 : method and theory , ed . joseph h . greenberg , charle a . ferguson and edith a . moravcsik , 123-156 . stanford : stanford university press . croft , william . 1990 . typology and universal . cambridge : cambridge university press . - . 1991 . syntactic category and grammatical relation : the cognitive organization of information . chicago : university of chicago press . dryer , matthew . 1989a . plural word . linguistic 27 . 865-95 . - . 1989b . article - noun order . cls 25 . 84-97 . - . 1989c . large linguistic area and language sample . study in language 13 . 257-92 . - . 1992 . the greenbergian word order correlation . language 68 : 81 - 138 . hopper , paul and sandra a . thompson . 1980 . transitivity in grammar and discourse . language 56 . 251-299 . koptjevskaja - tamm , maria . 1988 . a typology of action nominal construction . stockholm : university of stockholm . perkin , revere d . 1989 . statistical technique for determine language sample size . study in languagea013 : 293-315 . rijkhoff , jan , dik bakker , kee hengveld & peter kahrel . a method of language sample . study in language 17 . 169-203 . stassen , leon . 1985 . comparison and universal grammar . oxford : basil blackwell . tomlin , russell . 1986 . basic word order : functional principle . london : croom helm . dept of linguistic , u manchester , oxford rd , manchester m13 9pl , uk w . croft @ manchester . ac . uk fax : + 44-61 - 275 3187 phone : [phone]