3-445msg1.txt

贝叶斯学习算法源码

last thursday night , may 21 , zellig harri die in his sleep after a pleasant work day . he be 88 year old . he be bear in 1904 in byelorussium . i be tell that he choose the name zellig sabbettaus when his family immigrate to the unite state when he be four . i like to think that the semantics of happiness and steadfastness be on his mind . certainly they be keynote of his life . i would guess that his parent choose the name harri . when he die , he be just finish a book on politics that he have be plan for most of his life . with the 1992 publication of his book _ a theory of language and information _ ( oxford ) , he have wrap up his life 's work on language , at least for the time be . he seem to have feel at liberty to take up this other unfinish business . i understand from paul mattick , jr . , who be harri 's friend and neighbor for many year in new york , that this last book describe how to get from capitalism to socialism . this be surely not a conventional take on either capitalism or socialism , harri be an anarchist . oxford be interest in publish it , and he have also talk with cambridge . there be no memorial plan , beyond something very private for his family . however , there be some discussion begin of a public meet with scientific content . i would hope that the festschrift that haj ross call for in the lsa meeting some year ago may at last come into be . harri describe himself as a methodologist rather than a linguist . this could be mislead . he alway say that his work be not part of linguistics as it be institutionally define , and that linguist would not be interest in his work , though people interest in language would be . nonetheless , he be surely a linguist by most of the operational definition one may come up with . he have do extensive fieldwork on a variety of language . when he be do the final revision of the 1992 oxford book , he undertake to test the theory of language against every language of which he have some control , 44 language . he spend month read grammar from morn to night , and evaluate whether his theory have a reasonable account for what he find there . he be clear that no scientific conclusion be warrant , and so no particular notice of this check be give in the book , but he want to feel reasonably secure that his conclusion be not idiosyncratic to english , french , german , korean , and the few other language that have be the primary base for their development . he be please with the result . hbe contribution to the field be numerous and weighty . he found the first linguistics department in the u . s . he introduce the algebraic representation and abstract mathematical treatment which have become so much norm of the field that it be difficult now to appreciate how much he do so over the kick and scream protest of his peer . he invent x - bar notation for immediate constituent analysis , though of course not by that name , to cope with the well know weakness of ic analysis with the head-of relation . he develop way to accomodate discontinuous morpheme into grammatical analysis . he chart a way out of difficulty experience by bloch and other in phonology , by say that contrast rather than phonetic identity be the basis for set up phoneme , a ghost that have rise to haunt generative phonology more than once . he invent string analysis as a complement ( not rival ) to immediate constituent analysis . their complementarity with respect to the head-of problem be the basis of joshus 's tree - adjoin grammar ( tags ) . he invent transformational analysis in context of develop discourse analysis to get at the information content of text . other contribution await recognition and exploitation in the field of linguistics as institutionalize today , and in other field . obvious example include sublanguage analysis and sublanguage grammar , operator grammar base on word dependency , discourse analysis for information content , and his theory of information as an account of a central aspect of semantics . for example , string grammar and its natural extension into transformational grammar be the basis of the very successful work of naomus sager and other at nyu in information format of sublanguage text , apply there mainly to medical informatics . stephen johnson have implement a system for represent the information content of text , base on operator grammar . success of this sort be little notice within linguistics . it be characteristic of harri that there be no vanity or self importance in him . he know that his work be of last importance , and treat it as such , but he be no guru or empire builder seek follower , and would not accept any such role be project onto him . those student who seek entree to linguistics as a social institution in academia be bind to be disappoint . however , he could scarcely be blame for their disappointment . he do not provide such entree , nor do he pretend to , and in my hear actively discourage student who imagine work with him would further their ambition in the field . once , in my role as ta for john fought , i prepare a lecture on harri 's approach to syntax and semantics . as we be set out for the lecture hall , we encounter harri , and i blurt out " i ' m about to give a lecture on your theory to john 's class . " ( john , with characteristic wry humor , ask if he want to take anything back . ) harri bemusedly question whether anyone would be interest in what he be do . nonetheless , when he give a public lecture on " the two structure of language : report and paraphrase " in 1969 or 1970 , the large auditorium ( i think it be in the furness build ) be fill to capacity , and the critique by john corcoran , publish later in the volume _ transformationelle analyse _ edit by senta ploetz , be also well attend . broad attendance on and acclaim for his work could easily have be his , have he choose it . that be simply not where his ambition lay . a clue as to the basis of this choice against fame and influence may perhap be find in his advice to a student start out in his first teach position , many year ago . do n't invite anybody over for dinner , he say , and do n't accept any invitation . if you get involve in the social life of an academic , you win be able to get any work do . the work come first . harri be alway an intensely loyal man to his friend and family . the consequence , when combine with his laissez-faire anarchism , be not alway happy . hbe friend and close colleague of many year , henry hiz , be much more concern with build a formal linguistic program as an institution . the disparity of character could be devastate to student . i study with harri from 1966 through 1970 . i be an undergraduate much of that time , but that do not matter to him . he have a sink-or - swim approach like that attribute to sapir ( darnell 1990 ) , except that his seminar be of course focus on theory rather than the datum of , say , athabaskan . he would come in to his seminar and just start talk about what he be work on . when i start with him , this be the work that result in his 1968 book , _ mathematical structure of language _ ( wiley ) . the process be not a lecture or monologue , but a continue conversation with his student , try out alternative , pose and work out problem for a mathematical characterization of language . after a while , with intensive read outside , one begin to catch on and to participate . i recall tell him at the end of one seminar meet in my first year that i would try to disprove his theory . this trouble him not a bite . i work up a problem in modern greek that i think may be troublesome for his approach . ( i have live in greece for a couple of year , and speak the language , but i work with an informant for this project . ) when my result turn out actually to corroborate the point i have intend to challenge , he merely thank me for the datum on greek . a year or two later , i have come up with a proposal to analyze definition in a dictionary to extract semantic primitive by a form of componential analysis , much as martha even and now other have do . although the notion of semantic feature seem inimical in concept and method to his work , he say ( and this be an exact quote ) " other have try this and have fail , but you be welcome to try . " i offer this in refutation of the sometime hear view that harri be dictatorial . i run into conflict in such matter with hiz , never with harri . i have also hear it ask why he never retort to attack on his work . i think it do not matter to him . he do not expect his method and result to be understand and take up by everyone in the field of linguistics . maybe his attitude differ in the 1940 , when he write the structural restatement and the manuscript eventually publish as _ method in structural linguistic _ . ( btw , the title be to have say " descriptive " but the publisher substitute the buzzword " structural . " i recall him say , amusedly , " i do n't remember whether they ask me or not . " ) maybe his expectation of the field change after some of chomsky 's follower begin make him out to be the bid guy . i do n't think so , base on his writings and on the testimony of some who be his student then . i never hear him comment on the commonplace attribution to chomsky of the discovery of transformational grammar and the " transformational revolution . " there be a passage in _ the state of the art _ ( 1968 ) in which hockett attribute to harri " nothing , or a long silence , after 1957 , " show ignorance not only of thing like string analysis , for which he may be excuse , but even ignore the 1965 paper transformational theory prominently publish in _ language _ . i show this passage to harri , and he shrug . it do not matter . in particular , i never see any evidence that harri oppose or block chomsky 's ambition . in my experience it would have be entirely out of character for him . for example , it be harri who propose chomsky to speak in his stead to the 1962 international congress . a similar canard regard bernard bloch have recently be lay to rest in an editorial in language . one must i think be alert to the social psychology that lead some people to rewrite history so that their avatar be depict as an embattle hero . now , an old indian friend once tell me that one cannot point a finger without have three other finger of the same hand point back , so i hasten to add that this be not the picture i intend to paint here of harri . he accomplish what he intend to quite well , thank you very much , and seem to have be quite happy in the process . the point be precisely that he seem in no way embattle by attack and uncomprehend misconstrual of his work . and uncomprehend misconstrual abound . frawley 's review of _ a grammar of english on mathematical principle _ ( gemp ) be a good example . he identify harri 's operator grammar with predicate calculus , though harri be at pain to delineate critical difference between language ( a fortiorus operator grammar ) and language-like mathematical system , include predicate calculus . frawley can see in this comprehensive grammar only an attempt to do 1960 generative grammar in 1980 , because he be unable to step out of the generativist paradigm so as to understand harri 's work on its own term . another review ( eric wheeler , 1984 in _ computer in the humanity _ ) assert that harri 's grammar be unable to account for certain familiar semantic problem - - middle voice , the semantics of find v . seek , and quantifier scope in example like " someone be oppose by everyone . " in my review ( _ computational linguistic _ in 1984 ) i show how harri do in fact account of each of these problem in the book . michael kac , in his review of harri 's select writings , ask " why bother ? " and indeed , from within the generativist paradigm that must be the only plausible question . it be only in set aside paradigmatic blinkers that one can see , have these writings in one place , how consistent and self coherent harri 's program have be over the year . transformational grammar be not a revolutionary break but part of a continuous evolution . i will mention only one other misconception about harri 's work , not because it be in any way fundamental but because it be so commonplace . i probably will be greet with disbelief when i say that discovery procedure be not his aim . ( jim mccawley 's witticism about harri and discovery procedure in the collection traditionally circulate in may really reverse the role of the teller and the butt of the joke . ) it be not hard to see how linguist have come to this mistake belief . discovery procedure be an abide fixture for linguistics as institutionally define . when _ method _ be publish , linguist seek an aid to fieldwork and write of linguistic description . now , discovery procedure be institutionalize as a whip boy . this have color perception of harri 's intention and result . for harri , it be certainly of interest and value when redundancy on one level of linguistic representation could be use in a practical way to determine boundary of object on the next , but this be a corroborative byproduct , not an aim . the " constructional procedure " describe in the 1955 paper from phoneme to morpheme be implement in fortran in the early 1960 and prove to work , and ralph grishman have have some preliminary success in implement program to discover word class and rule of sublanguage grammar from sublanguage text . but in general harri do not think that discovery procedure be feasible . in particular , he tell me he think that grammatical analysis could not be do solely with a corpus or by ask informant , one have to control the language oneself . and then one have to work over the datum to tease out pattern and wrestle it into coherent form , a lengthy and demand process , as probably most of us know from experience . so much for the popularize image of feed in a corpus , turn a crank , and have a grammar reel out the other end . in the introduction to _ method in structural linguistic _ , harri state clearly that these method be not discovery procedure . he accept that one use many means to come up with proposal for describe what be go on in a language - - hunch , guess , heuristic rule of thumb , typological generalization , propose universal , comparison with relate language or earlier stage of the language , and so on , more art than science ( or rather , more art than engineer ) . harri be acutely aware of the danger of swamp one 's control of the language by grow familiarity with marginal example . language be after all a social institution , continuously in change as it be constantly recreate in the crucible of use . the aim of the method be not to substitute for these informal way of come up with possible analysis , but to verify , for any give result , whether the result have a valid relation to the datum of the language . of those who have actually read the book , how many have say ( and some have in fact say to me ) " he do n't really mean that . " but if nothing else , harri be alway careful to say exactly what he mean . this concern for verification arise out of a deeper concern which become more explicit in harri 's later work . this be a critical point for linguistics . for any other science , there be a standpoint external to the science domain for its metascience . in particular , practitioner in physics , chemistry , even in mathematics , rely on the " background vernacular " of language to ensure communication about share meaning and ultimately to validate the relation of conclusion , however reach , to the observation on which they be base . not so for a science of language . harri recognize and accept that there be no vantage point outside of language from which to describe language . and , observably , each language contain its own metalanguage . i ' ll repeat that , because it be i think a key to understand what harri be about , and because it be easy to overlook its importance . there be no vantage point outside of language from which to describe language . by contrast , generativist theory postulate a universal metalanguage , external to language , that be part of one 's biological endowment . ( i personally do not find this biologicist , neophrenologist doctrine of mental organ credible , but the issue rest not on opinion but on fact yet to be determine . ) this stance seem to me perfectly consonant with the argument make by stephen anderson in " why phonology be n't natural . " one cannot derive linguistic structure from the finding of some study bear a metascience relation to linguistics . harri be interest in how language can carry or transmit information , and this be the thread that underlie the really remarkable consistency in his work over more than 50 year . intuitively , we know that difference in form correlate with difference in mean , but the correlation be messy and inconsistent in the observe datum of language ( say , in a body of writings or of phonemic transcription , include whatever utterance the investigator may come up with in the ad hoc search for example ) . what harri find be how this messy , inconsistent stream of word can be the product of two concurrent system : a system of word dependency that correlate with perception in a subject-matter domain such as a science subfield , and a system of reduction that change word shape ( often to zero ) , motivate in part by issue of redundancy and efficiency and in part by historically contingent social convention . the reduction introduce degeneracy such as ambiguity and paraphrase , and otherwise obscure the correlation of form with mean , but without destroy that correlation . given that structure ( difference of form ) correlate with mean , it be of critical importance that the machinery of description not import any structure extraneous to that find in language . harri 's endeavor be alway , then , to determine a " least grammar , " a description that require an absolute minimum of primitive object and relation . any additional object or relation in the description introduce extrinsic structure that obscure the informational structure in language . this could be the basis for a tell critique of various other theory of language . harri choose not to make such a critique . when i ask him once about certain aspect of generativist theory , he would only comment , with evidence of mild amusement , that it do seem to be over-structure . like his teacher , sapir , harri have an interest in problem of international communication and an international auxiliary language . ( a paper on this appear in a 1962 volume on avoid world war iii . remember wwiii , everyone ? ) and like sapir and bloomfield he have in particular a long stand interest in international cooperation and communication in science . this culminate in _ the form of information in science : analysis of an immunology sublanguage _ ( with michael gottfry , tom ryckman , and other , 1989 , john benjamin ) . this book describe the grammar of the sublanguage of immunology during a specific period in the development of that field , base on discourse analysis of sublanguage text from that period and adequate for make explicit the information structure in arbitrary other text in that sublanguage . the analysis show how the structure of the sublanguage change concurrently with a change in immunologist ' perception in the domain of their science . a difference in informational structure correlate with a difference in mean . the informational structure that be clearly represent in the binary array result from discourse analysis be still present in the actual form of the source text as write albeit obscure under reduction in word shape , some motivate by consideration of informational efficiency and avoidance of redundancy , some dictate by convention of language use as a human social institution . harri arrange his life so as to enhance the autonomy of his work . i understand that his kibbutz in israel be a wealthy one , to which member give their asset and income , and which in turn support them in their need . i believe that the kibbutz purchase his apartment build on charle street . until his retirement , he hold an endow chair at penn , the benjamin franklin professorship in linguistics . he be principal investigator for a long series of grant from the nsf , nimh , and other agency whose committee and referee find his work of continue value . throughout his life he be involve with scientist and science . hbe wife be a physicist at the university of jerusalem , and have be albert einstein 's assistant at princeton . a brother be an immunologist ( he be an author of some of the work analyze in the 1989 book ) . he feel that the rough and tumble of polemic attack and retort be inappropriate for science , and would not participate in it . that too would be a distraction from the work . after one of the bampton lecture at columbium in 1986 , a young member of the audience approach him and ask what he would take up if he have another lifetime before him . he mention poetry , especially the longer work of 19th century poet like brown . he mention music . and he mention sign language . he have a long and very productive life . he have bring his life 's work to a successful culmination . with the completion of his book on politics , i imagine death come to him , as to the chess play knight in the seventh seal , and him say " ok , i ' m ready now . " it be a privilege to know him and to learn from him . he be an abide inspiration . bruce nevin bn @ bbn . com