Current Commen@S

Current Commen@s

Reflections on Learn@ Foreign Languages Part II: Electronic Translators

dumber 38 September 17,1979

In Part I of thw essay I dkcussed some demonstrated last year with “Speak and of the many language-teachhg courses, Spell,” a “toy” that seUs for $50.00. This methods, and devices available to per- toy contains a complete alphabetical sons interested in language learning. I keyboard, a visual display, and a digi- also expressed the hope that modern tized speech synthesizer. electronic technology may someday Out of curiosity, and hope that they make language learning easier. The re- might prove useful, I eagerly bought the cen! appearance of several new elec- Lexicon and Craig models when they tronic translating devices on the market first appeared. These devices are pre- is a significant step in that direction. sumably designed to make the job of In a future essay I intend to review translation easier. Unfortunately, that is the field of mechanical translation. My debatable. Portable, hand-held devices interest in that field goes back to 1954 somewhat larger than the average mini- when I first contemplated working calculator, they “translate” words and under Leon Dostert at Georgetown Uni- phrases from one language to another. versity. I My first contact in the field of They are designed to accept various information retrieval was James W. modules (or capsules) which provide the Perry, the author of the first self-teach- “memory” needed for translation. It ing book on scientific Russian.z Subse- would be inaccurate to say they are quently, I studied linguistics and the “programmed, ” as they only store inf or- problem of machine-translation of mation. chemical nomenclature.J It is tempting to compare these de- It is only natural, therefore, that I was vices to a portable dictionary. It is also excited by the appearance of several relevant to mention that nobody new translating devices. Less than a bothers to publish dictionaries that are year ago, Lexicon Corporation of limited to 15(N words or phrases. If they Florida introduced the first model, the did then maybe you could make a prop- Lexicon LK3C00. Craig Corporation of er comparison. But the fact is that most Japan followed scmn after with the Craig people cannot look up words in a dic- M1OO. Most recently, Texas Instru- tionary as quickly as they can with a ments (TI) announced the development keyboard. Furthermore, if they have to of a pronouncing translator. This will look up many words in succession combine the technology they developed frustration sets in. This is similar to the for calculators with the new technology point I made in discussing the problem of digitized speech. This was fmt of Russian-English dictionaries. 4

275 So in critiquing these devices one sonal experience. In 1960 I was a guest must remember that for many people lecturer at the National University of they will be much better than an inex- Mexico Computing Center. The direc- pensive pocket dictionary because tor, Sergio Beltran ,was interested in the many people simply find the use of dic- then very active field of mechanical tionaries difficult. To use either the Lex- translation. He wanted to develop a icon or Craig you simply key in, letter computerized system for translating by letter, the word you want to trans- English scientific texts into Spanish. So late. As you do so each letter is dis- he asked me to lecture his students on played on the screen. When you hit the mechanical translation and information “translate” key the translation is retrieval. He also insisted that I lecture displayed—quickly, but not necessarily in Spanish even though I protested that instantly. Except for certain specialized I had not spoken Spanish for many functions, the Craig and Lexicon oper- years. ate similarly. Both have a keyboard in So I went to the library. The only dic- which the letters are arranged in alpha- tionaries available were rather old. But I betical order. Other keys perform spac- managed and created technical words ing, backspacing, clearing, translation, whenever necessary. I wrote out my lec- and other functions. Unfortunately, ture and read it to the class. In it I sug- many translation requests result in the gested that the use of mechanical trans- display: “NOT FOUND,” (Lexicon) or lation as a substitute for learning ????? (Craig). Since the dictionaries only English was somewhat impractical. store between one and two thousand When my lecture was over, the students words you are bound to ask for words applauded politely. Then Sergio arose, that they do not contain. Before these thanked me for my lecture and said to devices can effectively compete with or- the audience, “YOU see my friends, dinary printed dictionaries they will mechanical translation does work. ” need to store at least 5,000 to 10,000 Electronic translators in their present words. There is every reason to believe stage of development would not have that as chip technology advances this aided me in my lecture. Nor will they will be achieved. In the meantime, un- enable you to engage in meaningful con- less it is a language you have never versations unless you have a lot of time studied, you are apt to be disappointed. and energy. Undoubtedly, under cer- In a recent review, Consumer Reports tain circumstances they will facilitate shared my opinion. They tested the Lex- communication. If you are trying to say icon, and although they felt that the something to a person who doesn’t translator was a promising beginning, in know your language, the visual display its present stage of development they of a phrase or sentence will have more felt it “would not be of much practical impact than pointing to the individual use for innocents abroad. ” They advised words in a dictionary. The device will the traveler to “take a good pocket dic- usually impress a person more than a tionary or foreign-language phrase book few scribbled words on a piece of paper instead. ”5 or some mispronounced words. Using one of these electronic transla- I pulled out my Craig translator in a tors will give you first-hand experience posh Paris hotel recently. The service of the limitations of word-for-word was disrupted for the next fifteen translation. Thk reminds me of a per- minutes. I had to lend it to the head

276 waiter so that we could continue our words, while it contains other words meal. Buthewas much more impressed, whose usefulness is questionable. as were my friends, by TI’s’’Speak and Neither the French nor German mod- Spell.” The impact of the spoken word ules, which I have, contains, for exam- is enormous. That’s why I suggest you ple, “have, ” “are,” “am,” or “hungry,” wait until TI comes out with its transla- although they do have “bourbon, ” tor. “scotch, ” and “rum. ” The manufacturer If you can live with such translations apparently never heard of the Thorn- as “the meat is able but the alcohol is dike word Iist,b a compilation of 30,G00 feeble” (the flesh is willing but the spirit English words ranked by frequency of is weak), then you can manage to create appearance in newspaper and other some useful sentences. When I showed commonly read texts. the device to a French neurologist she Many other words, such as “watch,” said it would be extremely useful for which have more than one meaning, are certain types of patients who could not programmed in the Lexicon with only speak. But apart from this, playing with one of their meanings. So when I tried the translator can bean enjoyable lear- to translate into French the phrase ningexperience. For example, by keying “Watch the movie” I got the equivalent in one, or several letters, such as “EV, ” of “Wristwatch the movie “ (Montre le you can easily display all the words that tine’ma). Or, when I tried to say, “Her begin with those letters. Certain phrases coat is there,” I got “Elle manteau est can also be accessed. If you first key in 12’—” She coat is there.” These par- “how,” a set of phrases such as “how are ticular difficulties, however, can be you, “ “how many,” and “how much, ” reduced if multiple meanings for hom- are successively dkplayed. To translate onyms are stored. any of these phrases, you press the The Craig translator does just that. It translation key. seems to have a wider, more complete The Craig has also stored 52 special vocabulary than the Lexicon. It includes phrases that can be retrieved by such common words as “have, ” “has,” touching a specified key and the “be,” “are,” and “am.” Words with two “Phrase” key. For example, “Do you meanings, like watch, are stored with change traveler’s checks?” can be ob- both meanings. When you key in such tained simply by pressing the “Phrase” words, they are “questioned” immedia- key followed by the letter V. The list of tely. You select the proper meaning common phrases is printed on the back before continuing. of the translator. The Craig also con- Neither translator, unfortunately, has tains categories of words, such as much grammatical sophistication. One “clothing, ” and “transportation. ” example of this problem in both devices Beyond their basic operational simi- is the treatment of verbs. The transla- larities, however, the Craig and Lexicon tors cannot conjugate. Most verbs are differ in many features. And, despite stored in their infinitive forms. So “I the fact that both are promoted as a want” becomes “Je dt%.irer” (Lexicon) or combination “dictionary, phrase book, “Je vouloir” (Craig)-in French, literal- teaching aid, and interpreter,” neither ly, “1 to want.” Try to say, for example, quite equals that blllkg. The Lexicon is “I want to eat ,“ and you get the transla- especially problem-ridden. It lacks tion “Je vouloir i manger,” “I to want to some of the most obvious and necessary to eat.” I suppose the average waiter

277 . will get the message. In addition, Despite their various problems, these neither translator has many past tense translators are proving interesting, if not verb forms, any verb-subject agree- especially useful. Since their introduc- ment, or any noun-subject agreement. tion, less than a year ago, over 300,000 If you spell poorly you will probably units have been sold. As a result, other have trouble with the Lexicon. When companies are entering this fast-grow- you spell a word wrong, you simply get ing field. T1 recently announced its new “NOT FOUND” displayed. Unless you translator. There are unconfirmed have a dictionary handy, or are good at reports that several other major com- devising alternative spellings, thk can panies will soon enter the translator be extremely frustrating. There is no market. T So far, prices are beyond the easy way to find out just why the word is reach of the student or average in- “NOT FOUND,” either. It may be mis- dividual. The TI translator is expected spelled, as mentioned, or may be to be priced at about $300, plus $50 for “Found” in only another form. (“See” is each module. The Lexicon currently found, but not “Saw”), or, of course, costs $225, plus $65 per module. The the word may simply not be in the Lex- Craig sells for $200, plus $25 per icon’s memory. Unlike the Lexicon, module. Both the Craig and the Lexicon however, when you misspell a word, or come with one module included. Al- type in a word not contained in the though I don’t have all of them, Craig’s memory, it flashes ????? on the modules for French, German, Spanish, screen. Pressing the “Search” key calls and Italian are currently produced for up the dictionary for all words spelled both models. similarly. So while you won’t find In addition, Lexicon produces Por- “worked” in the Craig’s memory, you tuguese and Greek modules, and a mul- will find, with ease, that “work” is ti-lingual module, which contains every- available. day phrases in English and five major Should you, for example, enter “sug- European languages. The manufacturer est” for translation, the Craig will re- promises that modules in Swedish, spond with ???????. Press the search Polish, Russian, Arabic, Hebrew, key, and the Craig begins to search its Japanese, and Chinese will soon be memory for alf words listed under first available. For languages that do not use “sugest, ” then “suges,” “suge,” and the Roman alphabet, a special keyboard finally “sug,” where it finds the entries overlay allows the user to enter the cor- “sugar” and “suggest ,“ which it displays. rect characters. Portuguese, Dutch, You then press the translation key for Russian, and Japanese modules are the translation. planned for the Craig translator. Unlike Since the Lexicon and Craig key- the Lexicon, the Craig has the capacity boards are arranged alphabetically, to translate to and from three languages anyone used to a typewriter keyboard, at one time. This requires inserting will have trouble. I suppose I would three modules at once. have preferred a standard typewriter ar- To a certain extent, of course, these rangement. To build up any speed key- devices rely on a process of transMera- ing in words takes some practice. It tion. I find this an exciting develop- might make some sense to locate all ment, as I have experienced the power vowels at the center top of the key- of transliteration several times. When I board. first lectured in Japan, many years ago,

278 I asked my friend Takashi Yamakawa, Even though they provide no gram- IS~’s first representative in Japan, to matical subtleties, or pronunciation translate into Japanese a short speech guidance, they can be helpful. The which I had prepared in English. Then I Chinese say that a trip of 1,000 miles transliterated his translation into begins with a few steps. If true, these Roman script. I read the speech to a devices may be the first steps toward the group of pharmaceutical industry scien- exciting “Universal Translator” used on tists in Osaka. They listened politely, the old Star Trek TV series.s That occasionally smiling at my pronuncia- translator, which was no larger than a tion, but in the end I received thunder- transistor radio, was able to ittstan- ous applause. That experience con- taneously convert alien languages to vinced me of the power of translitera- English and vice versa, Such a device tion. I had a similar experience when would, of course, be extremely useful. 1S1 was visited by a delegation of Until its creation, however, we will have Chinese scientists. With the help of an to use tradhional language learning 1S1 staff member I was able to say a few methods. When TI comes out with its words of greeting in Chinese. speaking translators, sociologists may Even though they are marvelous want to study their impact in certain technological feats, the electronic situations. translators are not substitutes for actual- The famous information scientist and ly learning languages. It is difficult to philosopher Joshua Bar-Hillel said that judge their value unless you use them in mechanical translation would never be a situation where you have no familiari- successful. I wonder what hls reaction ty with the language being translated. to these gadgets would be.

009)9 1s(

REFERENCES

1. Dostert L. Machine translation and automatic language data processing. (Howerton P W, ed. ) Vistas in information handling. Washington, DC: Spartan Books, 1%3. p, 92-110. 2, Perry I W. Scient~ic Russian. New York: Wiley InterScience, 1950.816 p. 3. Garfield E. Citation analysis, mechanical translation of chemical nomenclature, and the microstructure of science. Current Contents (5):5-8, 2 February 1976. (Reprinted in: Garffeld E. Essays of an information scientist. Philadelphia: 1S1 Press, 1977. Vol. 2, p. 415-8.) 4. ------Introducing ISI’S transliterated dictionary of the Russian language. Current Contents (14):5-8, 2 April 1979. 5. Is an electronic translator a traveler’s aid? Consumer Rep. 44:446-7, 1979. 6. Thorndfke E L & Lorge 1. Teacherk word book of thirty thousand words. New York: Columbia University Press, 1944.274 p. 7. Schuyten P J. Translator adds speech. NY Times 22 May 1979, p. D4. 8. Wbftffeld S E & Roddenbssry G. The making of Star Trek. New York: Ballantine Books, 1968.414 p.

279 February2, 1976 Number 5

The microstructure of science is very dif- All this is leading up to a discussion of ferent from its macrostructcsre. For some work I did which is rarely cited but example, I can confident y assert that which gave me fantastic satisfaction. I refer “milestone” papers-those which are sub- to a paper on mechanical translation of jectively rated as important by a large chemical nomenclature. 8 This was the sub- number of scientists-are, on average, fre- ject of my doctoral dissertation. Since 1‘m quently cited. However, 1 cannot truthful- so often asked why, 1‘d like to tell you how ly assert that every single“milestone” pa- I happened to take a degree in linguistics per is highly cited. A f- may have been rather than library science. almost totally ignored, for a variety of rea- I entered the field of documentation, sons. In fact, some portions of my own now information science, from chemistry work that I regard most highly have been by joining the Johns Hopkins University least cited. Thus it is painfully apparent to Indexing Project in 1951. I stayed until its me that models which are valid and relia- demise in 1953. By the middle of 1954 I ble in the macrostructurc of science can had already accumulated a msstcr’s degree crumble when the focus is narrowed to the in library science and sufficient graduate microstructure of wiencc. credits to satisfy the minimum require- How extensive this phenomenon is wc ments for a Ph.D. But it proved impossible have not yet been able to determine. I for mc to find a faculty member at Colum- wonder whether Watson and Crick would bia University who would approve for my agree that thcit 19>3 paper in Nuttire3 re- dissertation topic the use of machine me- presents the pinnacle of their work? 1 know thods in scientific information, The only that Oliver Lowry’f correctly asserts that his sympathetic ear was that of Professor most important papers ate not his most Merrell Flood, but in order to take a degree cited. But that does not say his most im- with him, 1 would have had to take under- portant are not heavily cited. Keep in graduate training in industrial engineering. mind that 1 am not saying citation analysis in retrospect, I see more clearly how rele- cannot detect the significant though infrc- vant systems work has been in my career. quendy cited pa~r. Back in 1964 we pro- I tried to form an interdisciplinary facul- duced a “computerized” history of DNA~ ty group, but I was not interested in which showed some papers that were infre- spending ten years trying to satisfy an in- quently cited but were significant in break- tetfacuky group that would supervise my ing the genetic code. work. By that time my family had already Examples of this kind have given me been convinced I was going to be a student reason to question the assertion by Co]& forever. I left Columbia disappointed. But that there is no validity in the Ortega hy- in 1954, through my friend and colleague, pothesis.7 This theory asserts that advances Casimir Borkowski, I met Professor Z.eliig in science depend in part on the contribu- Harris at the University of Pennsylvania, tions of mediocre scientists. While we may Deparrmcnt of Linguistics, His work in all stand on the shoulders of giants, they in structural linguistics was already well turn depend upon many average or less known to scholars, but in the field of sci- eminent scientists. Whether they depend entific information hc was unknown. In upon dwarfs is another question. 1956, I wrote a paper on the application of preprinted from J. C%em.Inform. & Comp. Sci. 15(3): 153-55,1975.

415 structural linguistics to mechanid indcx- Rccogniaing that the dictionary work a- ing9 and showed it to Harris. ‘fhottgh it Ionc might talcc me several yeats unless I wx never published, Harris became suf- got help, I proposed that the theory bc f~iencly interested in the field of irsforma- proven with respect to acyclic compounds. tion retrieval to accept some huge grants During the next few years 1 got into the from NSF over a ten-year period. Most of dcmiled problems of discourse analysis for this work is now continued ptimsuily by my target language-chemical nomcncla- Naomi Sager at New York University.10 turc. The details arc not essential to this Some of you may recall mutsformatiomd story. When I was ready for actual com- and diSCOUfSC atsdysis. puter trials, I got the help of John I suppose it was prestige that made me O’Connor in programming Univac 1, seek a Ph.D. I ultimatcl y worked out a which was then in use at Penn. But I found doctoral program with Professor Harris that I could never get time on the com- which commcnccd officially in 1958. Wc puter, so I had to buy time at the Franklin had agreed on the amount of course work Institute computer ccntcr. and my ulciiate d~rcatiocr topic. By The outcome of all this was ‘ ‘an algori- then 1 was quite preoccupied with prob- thm for translating chemical nomenclature Icms of chemical indexing. We were en- into molecular formulas. ” 12 When I sub- coding all new steroids for the U.S. patent mitted it to the department it was only tcn Ol%cc under a contract with the Pharma- pages. My substitute adviser was dumb- ceutical Manufacturers Association. founded by this. Dissertations in linguistics By 1960, the Institute for Scicntflc in- arc written by the pound-not the page. I formation (1S1) was publishing In&x spent a whole semester filling it out with Cbemtk. The original purpose of this ser- interesting theoretical statements and for- vice was to index compounds by molecular mal analyses of chemicaJ morphology, etc. formula. So it was natuml for me to want By late 1960 I had made the first succcssfd to find a way of calculating molecular for- computer run in calculating a molcctdar mulas in the simplest way possible. Until formula directly from a systematic name.8 that time everyone assumed that it was I had done this manually hundreds of necessary to dmw a stt-uctuml diagram in times ca.dicr in the yeax. order to calculate a molccukar formula. As it turned out 1S1 was never able to Even Aschcr Oplcr, 11 who wrote the pi- finance the research necessary to complete oneering paper in 1956 on “New Speed to this work. NSF was not very kindly dis- Structural Scarchcs’ ‘ , assumed this was the posed to us in those days. Wc also were up case. That is why hc fimt wanted to rcprc- to our cars in the G8rne#s2sCitation index scnt the compound in a topological matrix project so 1 had to put chemical nomen- which later was called a connectivity table, clature work on the back bumcr. Wc never My linguistic studies convinced mc that did input compound names for Cxme~t the ‘‘ mcaoing” of chemical nomenclature Ahtiats of Chwi.rW (CAC); on the con- had to include enough information for cal- trary, wc now input Wiswcsser Line Nota- culating molecular formuhu straight away. tion (WIN) for each compound and that is Otherwise, how could wc do this so quickly what wc use to compute the molecular in our heads for simple compounds? I tnld formula. However, the double bond check- Professor Harris my theory and hc accepted ing routines that wc used for so long were it as my doctoral thesis, the first in the ncw inchsdcd in my algorithm. field of chcmico-linguistics. Thanks to the About eight years ago 1 saw the proposal recognition by Professor Allen Day of cbemidAbstracts made to NSF regarding Penn’s Chemistry Department that it was a chemical nomenclature translation nontrivial problcm, the topic was agreed research. Naturally I felt envious that they upon in the gmduatc school. How~crt be- should get this support when it was clcady fore 1 could work on my dissertation, I had an opemtional development they needed to prove my theory worked. If it did not, I more than 1S1. That’s what made it ap- would have to choose another topic, no plied for thcm and academic for us. matter how long I spent on the research. However, I was very glad someone was

416 doing this and read with tied feelings complete. Consider that 25% of the ab- the fmtteports ofthis research in 1967.13 stracts in CA are of Russian material. 18 A recent paper in the Jouwd of Chemical From our extensive citation anal~s we Documentatios~4 shows chat this work is know that this is absurd in relation to the finally coming to fruition, and I congmcu- significance of Russian research. They are late the CA group on their accomplish- polluting the waters of science with a lot ment. of mediare and unrefereed material. Pro- Returning to the main point of my es- bably another 10% of CA falls into this say. Here is a topic of research which has category. No doubt others do it too, but multi-million dollar economic significance. the data show clearly chat the Russians arc There are only a few people in the world the worst offenders. Does anyone anywhere interested in it, so the number of times doubt the superiority of the Jowwd of tbe this kind of work will be cited is bound to Ametshm Chemud So&e/y over the Zhur- be small. Clearly it is the kind of thing nd Ohbcbei Kbimiz? How would you that is less cited than, e.g., papers on compare the abstracts of the ACS meetings WLN, but there is an important connect- to the abstracts of unpublished papers that ing thread. Perhaps historians will decide the Russians are now loading into the Rsi.r- that Opler’s notion of a connectivity table rtin Josmsd of Phyks6 Chemirby. Un- for chemical compounds has been the most doubtedly it gives the Russians sigti]cant important concept in this field. 11 Most political leverage to assert they account for people seem to think that Sussenguth was 25% of CA’s coverage. Maybe they will the first one to use his concept. 15 But even claim CA should pay them a royalty clearly none of these chemical information for abstracting without their permission. milestones has had any major discernible After all, CA abstracts do constitute a sub- impact outside the field, and that is what stimte for the original Russian material. the historian seeks and seems to find in There is an important distinction to be large-scale citation analyses. This again made between unrefereed material appear- demonstrates that the microstructure of sci- ing in high-priced journals and unrefereed ence is very different from its macrosrruc- material listed in a depository. Each ab- ture. stract requites the same space and work. So much for the history of mechanical But at least someone was willing to pay for translation of nomenclature. Let me digress that so-called high-priced journal. If libmr- now to make some observations on the ians are as indiscriminate as they are ac- future of chemical and scientific publica- cused of being, then why aren’t they buy- tion. This has been much in the news these ing the original Russian journals and ab- days, that is, C&E News! Joel Hildebmnd, stracts? I‘m sure that Earl Coleman would my freshman chemistry professor, has be delighted if libraries bought his tmns- caused a lot of soul-searching with his re- Iation journals without the slightest evalua- discoveryof the ancient idea of publication tion. He knows how hard it is to sell the by abstract. I’ve had some contact with best that the Russians publish. He would him in recent yearn and I know why he is court disaster to publish” everything with- making these proposals. Unlike James out regard to quality. Stemmie who in C&EN16 seems worried h is a rather interesting obwvation that that some important ideas will be lost to 10% of CA’s budget is about $2 million. posterity if we adopt any changed systems, If they cut back on Russian material they Hildebmnd is trying to tell us that the would find the same $2 million they want system is overloaded with useless informa- the Russians to pay for pirating CA. tion; he is talking about information pol- At 1S1we have very mixed feelings about lution on a huge scale. I have rceentlyll CA. On the one hand, wc resent theit high asserted that the abuse of the page-charge price because a chemistry department is system may be aggmvating this pollution generally apt to say that it can’t -afTordthe problem. And I regret to say Chemicaf Science Citatib# Index (SCI) but it must Abstrads may be qually guilty. CA does buy CA. If for no other reason, it couldn’t this unwittingly in its hopeless aim to be get ACS accreditation without it. On the

417 ...... ---- other hand, the higher CA’s price chcckmg rhe Items retmved m the WI! becomes, the more easily we can convince This is frequently done when people use buyers that SCI or CAC is a good value. MEDLINEand SCISEARCH,2but obvious- However, given my choice, I would much ly the inclination to do so is tempered by rather see CA priced lower. So I have a real the vast differences in per-hour rates. concern for their cost-effectiveness. In fact, In closing, I will mention miniprint, given my druthers, 1 would provide for CA which has now come into the limelight. As a citation index to the chemical literature the cost of paper goes up, CA and 1S1may that would complement CA searches. The well have to adopt such methods. Whether combined use of CA and SCI is happening users will accept miniprint more readily increasingly, but it would be nice if we than microform is hard to determine, but could accelerate the use of SCI by chemists there is a whole new technology opening as was suggested by the Hartnay Commit- up now that the ‘‘Oxf6rd English Diction- tee many years ago. 19 ary” has become so successful in this medi- The recent paper by Party, Linford, and um. Ralph Shaw and Albert Boni experi- Rich1 shows a clear trend toward such tom. mented with miniprint long ago. I just re- plcmcntary use of large data bases, This diwovcrcd it when I was thinking about will increase as the cost of on-line services ways to cut down on indexing costs. Maybe declines. it’s still not too late for CA to try it. After I recently did a search of the CA data all, the most successful publishing vemurc base using our Permutenw Subject InAx of the past dccadc has been in the mini- (PSI) to identifi pertinent search terms and print edition of the “Oxford English Dic- then followed up the output from CA by tionary”.

1. ParryAA, LirrfordRG&RichJ I. Com- 11. Opler A & Norton T R. New speed to puter Iiteratutc searches;a comparison of the stmctuml scarchcs. Chern. Eng. News performmcc of two commercial systems in an 34:2812-14, 1956. interdisciplinary subject. hrf SCi. 8:179-87, 12. Gartield E. An afgontbm for tradating 1974. cbemicd names to molecrdar formrdas. 2. Gar&ld E. 1S1’s SCISEARCH timc- (Philadelphia: Institute for scientific Informa- shaccd system trades time for money--but arc tion, 1961), 68 pp. you ready for this? Crmwr/ Cofimr@ 13. VanderStouw G G, Naznirsky I & (C@ ) No. 40,4 October 1972, p. 5-6. RushJ E. Proccdurcs for converting systematic 3. Watson J D & Crick F H C. A structure names of organic compounds into atom-bond for dcoxyribosc nucleic acid. Nuture 171:737, connection tables. J. C&em. Dec. 7: 165.69, 1953. 1967. 4. Lwvty O. Personal communication to D.J.D. Price, quoted in: Garficld E. Citation 14. VanderStouw G G, Elliott P M & frequency as a measure of research activity and Iaenberg A C. Automatic conversion of chem- performzncc. CC No. 5, 31 Jan’1973, p. 5-7. ical substance names to atom-bond connection tables. J C~esn. Dec. 14: 185-93, 1974. 5. Garfield E, Sher 1 H & Torpie R J. T4e use of citation data in wnh”rngthe hidosy of 15. Susenguth, E H. Gmph theoretic algor- science. (Philadelphia: Institute for Scientific ithm for matching chemical structures. J, Information, 1964), 86 pp. Cbem. Dec. 5:36-43, 1965. 6. Cole J R & Cole S. The Ottcga hypo- 16. StesnsnleJT. Control of scientific papers thesis. Science 178:368-75, 1972, Cbem. Eng. New~ 53:33-34, 1975. 7. Ortega y Gamer J. The revo/t of /be 17. Garfield E. Page charges; for profit and mmse$. (NCWYork: Norton, 1932), p. 84-85. non-profit journals; and freedom of the scien- 8. Garfield E. Chemico-linguistics; compu- tific press. CC No. 7, 17 February 1975, p. ter translation of chcmicai nomenclature. 5-7. N@ure 192:192, 1961. 18. Baker D. World’s chemical literature con- 9. Garfield E. Proposal for research in mechanical indexing. Unpublished manuscript, tinues to expand. CAem. Eng. News 49: 37-40, 1971. 1956. 10. Sager N. Syntactic formatting of scicncc 19. Anonymous. ACS report mtcs informa- information. AFIPS Conf Pmt. 41: 791-800, tion systcm cfficicnc y, CAem. E#g. Nesw 47: 1972. 45-46, 1969.

418