Báo cáo khoa học: " The Signal System in Interlingua — A Factor in Mechanical Translation"

Chia sẻ: Nghetay_1 Nghetay_1 | Ngày: | Loại File: PDF | Số trang:0

Thêm vào BST

Báo xấu

81
lượt xem 3
download

Download Vui lòng tải xuống để xem tài liệu đầy đủ

PREOCCUPATION with the problems of communication across language barriers has resulted in our time in the perfection, progressive application, or investigation of three new techniques. They are (1) simultaneous translation, as practiced in the United Nations General Assembly, and more and more extensively also in the most varied international congresses of scientists and other groups of specialized endeavor, (2) Interlingua, as utilized currently (especially for medical summaries) in increasing numbers of scientific periodicals and printed programs of international congresses, and (3) mechanical translation by electronic computers, as envisaged especially by scholars at Georgetown University, the Massachusetts Institute of Technology, the...

Chủ đề:

Bình luận(0) Đăng nhập để gửi bình luận!

Lưu

Nội dung Text: Báo cáo khoa học: " The Signal System in Interlingua — A Factor in Mechanical Translation"

[ Mechanical Translation, vol.2 no.3, December 1955; pp. 55-60] The Signal System in Interlingua — A Factor in Mechanical Translation Alexander Gode, Science Service, Division de Interlingua, New York City PREOCCUPATION with the problems of commu- of simultaneous translation need never and can nication across language barriers has resulted never serve as an "official" rendering ready and in our time in the perfection, progressive appli- valid for incorporation in a permanent printed or cation, or investigation of three new techniques. otherwise published record. They are (1) simultaneous translation, as prac- The relation of Interlingua to the technique of ticed in the United Nations General Assembly, simultaneous translation is on the whole one of and more and more extensively also in the most irrelation. There can be no competition between varied international congresses of scientists the two but only peaceful coexistence and -- let and other groups of specialized endeavor, (2) us hope -- cooperation. An ideal example of mu- Interlingua, as utilized currently (especially for tual complementation of these two techniques is medical summaries) in increasing numbers of that of the forthcoming Sixth Congress of the In- scientific periodicals and printed programs of ternational Society of Hematology. This Congress international congresses, and (3) mechanical has announced its selection of English, Spanish, translation by electronic computers, as envi- French, and Interlingua as official languages. saged especially by scholars at Georgetown Uni- All publications -- announcements, programs, versity, the Massachusetts Institute of Techno- etc. -- which the Congress sees fit to publish logy, the Universities of California and Washing- will be in either English and Interlingua or in ton, and Birkbeck College (University of London). Interlingua alone. All papers presented at the The technique of simultaneous translation Congress will be read in English, French, or seems definitely established. It is to be assumed Spanish with simultaneous translation being pro- that it can be further perfected by technological vided for these three languages. improvements of the machines with which it op- We may, if we wish, dream of a further sim - erates and also by additional refinements in the plification of this already highly efficient setup. training of the translating personnel. The psy- Instead of three languages admitted for the pre- chological processes which the simultaneous sentation of papers, we may wish for an ultimate translator experiences--his tendency to identify liberalization shedding all restrictions. Under himself with the orator after the fashion of a such a fantastically ideal setup, participants in good actor, his "schizophrenic" endeavor to be the Fiftieth or Seventy-Fifth International Hema- simultaneously, so to speak, a listening Russian tological Congress may present their papers in balalaika and a resonant Spanish guitar, and so Dutch, Hindustani, Japanese, Hungarian, Finnish, forth--make fascinating material for descriptive M arathi . . . or any other language of their and analytical studies in some branch of meta- choosing, with no chaos resulting, thanks to the linguistics. But no fundamental research seems technique of simultaneous translation and its pro- required at this time to lift the technique of si- vision of Interlingua versions of every individual multaneous translation into the saddle or to keep contribution. it there. Those who cannot appreciate the well-nigh There is likewise very little to be done to ex- ideal efficiency of these two plans (with the sec- tend the limits of the potential usefulness of si- ond thrown in to placate incorrigible pursuers multaneous translation. These limits are com- of the as yet unlikely), those, i.e., who hold that pletely clear, even if they represent a suit of no international congress can be said to have armor the little giant must grow a lot to fill out handled the language problem efficiently if it completely. Simultaneous translation is a tech- does not provide for one common auxiliary lan- nique that can be applied wherever the spoken guage for all participants, are looking at things word in one particular language needs to be un- from a viewpoint that is alien to me. I do not derstood for immediate reaction by groups of wish to criticize them, but I also do not wish to individuals whose language masteries do not be associated or confused with them. cover that one particular language. The product If it seems impossible to construe any sort of 55
56 A lexander Gode potential, let alone real, rivalry between the all the parts that are needed for an English trans- techniques of simultaneous translation and Inter- lation can be covered by simple equations, and lingua, the same obtains to an even more striking even the rendering of this particular "in" as 'into' degree for the techniques of simultaneous and obeys the objective demand of the following ac- electronic translation. cusative, "die Stadt." I could very well expect The suggestion that in some distant future it a machine to render the passage correctly as, may become possible to replace the simultaneous 'I go into the city." By a slightly more complex translator by an electronic computer which scru- system of equations I could also expect a ma- tinizes, analyzes, and then translates the spoken chine to translate, equally correctly, 'I go down- word, belongs as much in the realm of science town. ' But I do not see how a machine can recog- fiction as does the idea of one universal language nize which of the two, 'I go into the city' or 'I go used by all mankind in all transnational forms of downtown,' is to be preferred in a given context. communication. Or take an example culled from an actual medi- The juxtaposition of the techniques of simulta- cal text, where the American author speaks of neous translation and mechanical translation by 'atrial fibrillation and flutter.' Let us suppose electronic computers reveals one striking differ- we want a machine to translate this passage into ence. The simultaneous translator cannot and French. There are but four equations involved must not attempt to analyze. If he parses the which yield the French words, "auriculaire, fi- statement he hears and then proceeds to search brillation, et, flutter." And yet I do not see how for the best word, he is lost. He must grasp the a machine can decide whether the correct trans- orator's statements as phraseological Gestalten lation is "fibrillation auriculaire et flutter" or and react by the spontaneous production of cor- r ather "fibrillation et flutter auriculaires." responding Gestalten in the target language as- The point of interest here is not that we have signed to him. If there are intervening processes proved that machine translation must break down of analysis and resynthesis in the translator's under certain conditions. Perhaps we haven't. mind, the translator himself must not be aware The point of interest is that we think we have de- of them any more than a healthy diner is aware monstrated such a breakdown because we cannot of what happens to a slice of steak on its way to find an objective and unambiguous indicator or generating a new supply of red blood corpuscles. signal which decides that 'I go downtown' is cor- In contrast to all this the technique of mechani- rect and not 'I go into the city,' that "fibrillation cal translation presupposes the most careful and et flutter auriculaires" is right and "fibrillation the most detailed analysis. auriculaire et flutter" wrong. W e know -- and admit or emphasize -- that I n lieu of our earlier reference to the process t ranslation is an art. But much of it can be re- of translation as a complex activity reducible duced to simple equations which can be recorded with or without a remnant to objective equations, for future reference and which are always cor- we might say that the process of translation rect. How much of it? amounts to making, in the target language, state- In this question, I believe, we have in a nutshell ments which heed all the signals appearing in the whole problem of mechanical translation. If whatever we or a machine are trying to translate. not only much of the process of translation but In this global conception any spoken or written all of it could be reduced to verifiable and ever- passage consists of signals, nothing but signals. valid equations, the linguistic side of the problem These might be classified as semantic signals of mechanical translation would be solved; nothing ('cheese' is a semantic signal which suggests the would remain open and pending but the construc- entity 'putrid milk'), intonation signals (depend- tion of a mass-producible, economical machine. ing on its intonation, 'no' may signal surprise, I am sure, if Mr. Thomas Watson should fail to incredulity, or rejection), grammatical signals avail himself of this opportunity to make his name ("die" in "Ich gehe in die Stadt" signals a rela - completely immortal, the late Mr. Henry Ford tionship of movement into the following entity), would step into the breach. All of which is meant etc. to point up the fact that in the whole realm of me- For more conventional purposes it may be chanical translation the engineer is ready while better to restrict the meaning of the term 'signal' the linguist is not. more or less to what I have just designated as It may smack of prejudice if I insert at this 'grammatical signals.' In the present context, point the opinion that the complete reduction of however, we need the more comprehensive inter- the process of translation to objectively valid pretation. It permits us to expand an earlier al- equations seems impossible. If we consider the lusion and define the task of the researcher in simple German statement, "Ich gehe in die Stadt," mechanical translation as amounting to the elabo-
S ignal System in Interlingua 57 ration of a system whereby all the elements ap- 4. Is reported the failure of essays to in - pearing in the finished translation are unambigu- duce leukemia in young mouses feminine of the ously derived from objectively recognizable sig- stirps DBA2 by paint them with 20-methylchol- nals in the original. anthrene in benzene. This approach permits a type of experimenta- tion which brings out two important principles. The critique of this translation will suggest a First, the system of signals in any given language few improvements of word choice ('strain' in- can be described as consisting of various catego- stead of 'stirps,' 'female' instead of 'feminine,' ries. Second, the refinement of signal categories etc.); it will demand correct irregular plurals and sub-categories that need to be considered in ('mice' instead of 'mouses') and the use of the a given translation problem depends on the rela- lonely personal ending in the third person of the tion between the signal systems of the departure present tense; and finally it will point out as the language and the target language. In other words only major weakness of the translation the un- In a given pair of languages that are to be inter- English position of the adjective which overshad- related by the process of mechanical translation, ows all other blemishes (including the single the categories of signals need not be exhaustive. instance of a misplaced reflexive pronoun). If we interrelate two languages by such a process If we edit the translation in accordance with of translation, we can stipulate experimentally these observations (taking only one or two addi- that we want to heed only a specific set of signals, tional liberties of minor significance), we obtain The result of the translation effort can then be the following version. criticized in order to determine whether it could be improved by heeding additional signal catego- An in Vitro Study of Antileukemic Serum ries, how far the heeding of ever subtler catego- ries can perfect the finished translation, and 1. We prepared in rabbits an antiserum to whether there is really a remnant of indispensable lymphogenous leukemia induced in mice of the elements which the target language requires but strain DBA2, containing antibody against nor- which cannot be inferred from objective signals. mal and leukemic lymphocytic antigens in ac - I present a sample translation from Interlingua cordance with determinations by the complement into English in which in addition to all semantic fixation test of Thornton et al. equations only the signals for tenses, participles, 2. When this antiserum was incubated with and plurals are heeded. The passage was chosen normal lymphocytic antigen, all its complement- at random and happens to be the author's summary fixing activity was eliminated except that which of a medical paper. reacted with leukemic tissue. It seems that an antibody or a group of antibodies was produced A Study in Vitro of Serum Antileukemic that was specific for this leukemia. 1. Was prepared in rabbits a antiserum anti 3. An antiserum to lymphogenous leukemia leukemia lymphogenous induced in mouses of the induced in mice of the strain DBA2 did not show stirps DBA2, containing antibody against anti- itself able, either in prophylactic or therapeutic gens lymphocytic normal and leukemic accord- administrations, to protect to a significant degree ing to determinations by the test of fixation of other mice of the strain DBA2 which had received complement of Thornton et al. inoculations of transplantable leukemia of the same cellular type. 2. When this antiserum was incubated with antigen lymphocytic normal, all its activity com- 4. We report the failure of attempts to induce plement-fixing was eliminated except it which leukemia in young female mice of the strain DBA2 reacted with tissue leukemic. It seem that a by painting them with 20-methylcholanthrene in antibody or a group of antibodys was produced benzene. that was specific for this leukemia. 3. A antiserum anti a leukemia lymphoge- Aside from the question as to how much of the nous induced in mouses of the stirps DBA2 not editing performed on the above piece could be itself showed capable, so much in administra- reduced to mechanical reactions to signals in the tions prophylactic as also therapeutic, to pro- original, there is also the question whether the tect to degrees significant other mouses of the comparatively satisfactory result was not stirps DBA2. the which had received inoculations possibly due to a very high degree of kinship of leukemia transplantable of the same type cel- between the two languages involved, i.e., between lular. Interlingua and English.
58 Alexander Gode There can be no doubt about the closeness of the Interlingua passage itself may here be inserted kinship of Interlingua and English. But this kin- for the sake of completeness. ship is not exclusive; it is a consequence less of Un Studio in Vitro de Sero Antileucemic the nature of English than of Interlingua. 1. Esseva preparate in conilios un antisero I suspect that many of the implications and con- anti leucemia lymphogene inducite in muses del clusions of the present survey would collapse or stirpe DBA2, continente anticorpore contra an- would have to be modified if it were extended to tigenos lymphocytic normal e leucemic secundo the Slavonic and further to non-European lan- determinationes per le test de fixation de com- guages. That extension (and modification of my plemento de Thornton et al. doctrine) I have to leave to others who are qua- lified to tackle the problem. Left to my own 2. Quando iste antisero esseva incubate con devices, I can merely claim that an experimen- antigeno lymphocytic normal, omne su activitate tal juxtaposition of Interlingua with any other Eu- complemento-fixante esseva eliminate excepte ropean language, carried through after the fore- illo que reageva con texito leucemic. Il pare going Interlingua-English model, would yield the que un anticorpore o un gruppo de anticorpores same type of result. esseva producite que esseva specific pro iste The first passage of our Interlingua text, me- leucemia. chanically translated into German, would read: 3. Un antisero anti un leucemia lymphogene inducite in muses del stirpe DBA2 non se mon- War (wurde) bereitet in Kaninchen ein Anti- strava capace, tanto in administrationes prophy- serum anti Leukämie lymphogen induziert in l actic como etiam therapeutic, a proteger a Maus (pl.) von d- (der, die, etc.) Stamm DBA2, grades significative altere muses del stirpe DBA2 enthaltend Gegensubstanz gegen Antigen (pl.) le quales habeva recipite inoculationes de leuce lymphozytisch normal und leukämisch gemäss mia transplantabile del mesme typo cellular. Bestimmung (pl.) durch d- (der, die, etc.) Test von Fixierung von Komplement von Thornton et 4. Es reportate le fallimento de essayos a al. inducer leucemia in juvene muses feminin del The same passage in French would read: stirpe DBA2 per pinger los con 20-methylchol- Etais (était, etc.) préparé en lapins un anti- anthrena in benzina. sérum anti leucémie lymphogène induit en souris This Interlingua passage was obtained by the de l- (le, la, les) race DBA2, contenant anticorps devices of human, i.e., non-mechanical transla- contre antigènes lymphocytique normal et leu- tion from an English original which read: cémique selon déterminations par 1- (le, la, les) test de fixation de complément de Thornton et al. An in Vitro Study of Antileukemic Serum This French and the preceding German, no less 1. A rabbit anti-DBA2-mouse-induced lym- than the fuller English sample, are definitely phogenous leukemia serum was prepared that editable, i.e., if we suppose that a mechanical- contained antibodies to normal lymphocytic and translation setup could produce such texts on a to leukemic lymphocytic antigens, as determined large scale, MT (as the experts call mechanical by the complement fixation test of Thornton and translation) would be in business. One feels h is associates. tempted to assert that in the presence of an out- 2. When this antiserum was incubated with put of such quality, the question of whether a normal lymphocytic antigen, all of its comple- more refined heeding of existing signals can im- ment-fixing activity was removed except that prove the output, or perhaps actually make it which reacted with the leukemic tissue. It ap- perfect, recedes to a fairly academic background. pears that an antibody or group of antibodies In any event, the explanation of the compara- was produced which was specific for this leuke- tively high quality of our results lies in the spe- mia. cific character of the signal system of Interlin- 3. An antiserum to lymphogenous leukemia g ua in relation to that of English, French, induced in DBA2 mice, given prophylactically German, etc. or therapeutically, did not significantly protect It should be possible to dramatize this asser- other DBA2 mice that had been inoculated with tion experimentally by a mechanical translation a transplantable leukemia of the same cell type. interconnecting German, French, English or other languages not including Interlingua. Let us use 4. The failure to induce leukemia in young for this purpose the English text on which the DBA2 female mice by painting them with 20- above-used Interlingua passage was based. The methylcholanthrene in benzene is reported.
S ignal System in Interlingua 59 In putting this passage mechanically into French While the samples of English, German, and or German, our interest is to see whether the French evolved by mechanical translation from product is editable as the corresponding product an Interlingua starting point were so eminently based on Interlingua was editable. The French comprehensible and readily editable that a re- result is as follows: finement of the mechanical process lost at least some of its urgency, the German and French Un(e) in vitro étude de antileucémique serum samples evolved from an English base are at least in part so eminently incomprehensible and 1. Un(e) lapin anti-DBA2-souri-induit lym- uneditable that an immediate identification of the phogène leucémie sérum étais (était, etc.) pré- responsible factors becomes imperative. Let us parais (préparait, etc., préparé) que contenais take up at least one representative case. (contenait, etc., contenu) anticorps à normal lymphocytique et à leucémique lymphocytique What could we do to eliminate or reduce the antigène, comme déterminais (déterminait, etc., utter confusion of "un lapin anti-DBA2-souri- déterminé) par le complément fixation test de induit lymphogène leucémique sérum" and "ein Thornton et son (sa, ses) associés. Kaninchen anti-DBA2-Maus-induziert lympho- gen Leukämie Serum"? What additional signals 2. Quand ce (cet, etc.) antisérum étais (était, could we have heeded in the English original, etc.) incubé avec normal lymphocytique antigène, 'a rabbit anti-DBA2-mouse-induced lymphoge- tout (tous) son (sa, ses) complément fixant acti- nous leukemia serum' ? More specifically: What vité étais (était, etc.) éloignais (éloignait, etc., s ignals are there to decide whether this is a éloigné) excepté que (celui-là, etc.) que (celui- 'leukemia serum' which happens to be 'lympho- l à, etc.) réag-ais (réag-ait, etc., réag-é) avec genous' or a 'serum' of 'lymphogenous leukemia'? le (la, les) leucémique tissu. Il apparaît- que W hether it is a 'rabbit leukemia' or a 'rabbit se- u n(e) anticorps ou groupe de anticorps étais rum'? Whether it is a 'serum induced against (était, etc.) produi-ais (produi-ait, etc., produit) DBA2 mice' or a 'leukemia induced by anti-mice'? que étais (était, etc.) spécifique pour ce (cet,etc.) The fact of the matter is that there are no sig- leucémie. nals to answer these silly questions and quite a 3. Un(e) antisérum à lymphogénique leucémie few other less silly ones. The English passage induit en DBA2 souris, donne prophylactiquement is not grammatically comprehensible to anyone ou thérapeutiquement, fai-ais (fai-ait, etc.) ne- not specially prepared by information about the pas protéger autre DBA2 souris que (celui-là, subject matter. etc.) av-ais (av-ait) été inoculais (inoculait, etc., inoculé) avec transplantable leucémie de le (la, The passage under discussion may be extreme, les) même cellule type. but it is certainly not unrepresentative. English is rich in unsignaled relationships of a peculiarly 4. Le (la, les) faillite à induire leucémie en complex kind. But, the presence of unsignaled jeune DBA2 féminin souris par colorant les (eux) relationships in English or in any other language avec 20-methylcholanthrène en benzine est rap- is not especially noteworthy. It is rather the ab- porté. sence of such relationships that would be news, There are in this string of French words cer- and incredible news to boot. Signalwise, snow - t ain sequences that might make sense to an man, milkman, pitman are quite alike, yet we editor of good will. But there are others that know that a pitman is not a man made of pits; a cannot possibly be parsed by anyone unless snowman is not a man who sells snow; and a he knows English, and knows it at least well milkman is not a man who does his work sub- enough to tackle the translation without mecha- merged in milk, even though we have to gather nical help in the first place. The impression that knowledge from experience not reflected in left by the corresponding German product is not the corresponding word forms. Signalwise, "Ich much better. In lieu of the complete text, this gefalle ihm" and "Ich folge ihm" are quite alike, sample may illustrate the point: yet we know that the first statement involves a reaction on his part, the second an action on my Ein (eine, etc.) Kaninchen anti-DBA2-Maus- part although there is no objective signal to induziert(e) lymphogen Leukämie Serum war mark this difference. (wurde) bereitet(e) dass (das, der, etc.) enthal- What is important from the point of view of t-et(e) Gegensubstanz zu normal lymphozytisch translation and of mechanical translation in par- und zu leukämisch lymphozytisch Antigen (pl.), ticular is not that the signal system of departure wie bestimmt(e) durch d-(der, die, etc.) Kom- language and target language be complete in any plement Fixierung Test von Thornton und sein- absolute sense of the term but rather that they Sozius (pl.).
60 Alexander Gode agree that editable results are a first step toward be compatible. If the departure language supplies their more ambitious goal, and this enables them signals for categories which the target language to cooperate with those who hold that mechanical does not and cannot represent by special forms translation need not and should not aim at any- and leaves unsignaled other categories which the thing more than the production of editable texts. target language requires, the translation becomes I have attempted to show in this paper that a correspondingly more difficult and may even turn base text in Interlingua is convertible by mecha- out to be impossible. nical means into an editable translation in a In the case of the languages used for illustra- target language belonging to the group of lan- tive purposes in this paper such difficulties are guages which are summarized in Interlingua. not insuperable but they are quite real. In evolv- This does not imply that the same cannot be true ing texts in any of these languages from Inter- for languages outside that group. It merely im- lingua, however, they are all but non-existent. plies that such a more comprehensive assertion The reason for this seemingly surprising obser- requires additional experimentation by compe- vation is not hard to find. The categories for- tent investigators. mally signaled in Interlingua are those and only those which the languages summarized in it have In any event, there is a group of languages in common. If one of the base languages of Inter- (possibly quite extensive) which form a circle lingua dispenses with a particular signal cate- the center of which is occupied as it were by gory, so does Interlingua. If we translate mecha- Interlingua. This suggests the possibility of uti- nically from Interlingua into English or French lizing Interlingua in mechanical translation as or German or any other language of the same an intermediate language. A first step may have general group, we find of necessity that (1) Inter- to be a more precise determination of what lan- lingua gives no signals which our target language guages could be profitably involved in such a finds it impossible to utilize and (2) Interlingua's system. The second step would be the mechani- failure to supply signals of various types custo- zation of the translation of texts written in Inter- m arily present in the target language is re- lingua with all the links in its surrounding circle stricted to instances where comprehensibility as target languages. If as a third step the re- and hence editability is not impaired. verse process of translating into Interlingua were An English text which never signals the dif- likewise mechanized, all the languages in a group ference between ordinary and progressive tenses of n languages could be interconnected by 2n pro- may sound queer, but it is comprehensible and cesses of mechanical translation instead of by n2-n such processes. The linking of twenty lan- editable. The same goes for a German text which never signals the difference between a pronoun guages in all directions would not require three t hat refers to "der Tisch" and one that refers to hundred and eighty processes but only forty. "die Uhr" or "das Buch". And exactly the same, too, goes for a French text which never signals Acknowledgment.-- The author's interest in pro- by a verbal ending whether the first, second, or blems of MT was first aroused by Professor Léon third person is meant. Dostert of Georgetown University. It matured It is true that many of the specialists of me - through contact with Drs. Victor H. Yngve and chanical-translation research are not satisfied William N. Locke of MIT. It turned to actual re- with editable products. They evidently must search thanks to the encouragement and guidance have arguments which defeat everything I have received from Dr. Yehoshua Bar-Hillel, likewise said to show that there are translation situations of MIT. Indebtedness to others, too numerous to in which mechanical processes cannot possibly l ist, is herewith acknowledged in cordial grati- yield editable results, let alone results that re- tude. quire no editing whatever. Yet these men will