rnaJuurnrnaJlb L10lEV 0 BrnI UU aJ rn B slDlE s '1 00 SaJfil S f OOaJlE l1aJWfil Brn lluml1mrn t Jmrn $ t Jml1 0 1 EO 1 4 c P L 86-36 86-36 COMPUTERS IN EL INT AND TELEMETRy o1 I o o 1 RIGHT ON VERA o ooo o o o o oooooooooooo A o o o o o 7 Iooooo oo 8 TRANSLITERATION OR CyRILLIC oo o o COMMENTS ON AG-22 IATS ooo oo o o oooooo 13 NSA-CROSTIC No 4 oo ooo o oooo o A J S ooo o o o J 18 THEi MARQUI S AND THE MEDIUM o o ooooo Reed Dawson oo o ooo 20 THEI ICOLLECTION SYSTEM o o oooooo Tirn Murphy oo ooo 21 HOW THINGS HAVE CHANGED oo oo o o 1 l 25 CHINESE MACHINE TRANSLATION o NOTES ON TRANSLATION FROM THE CHINESE j 26 SCIENTIFIC CHINESE MACHINE TRANSLATION 28 LETTER TO THE EDITOR o oo oooo ooo o o o oo 29 t elwwiled b BIRNSi eneSB tNSi eSSM 129-11' Exempt riOlii BS EO 11852 eategiWy 2 ge__if p j eileMioo lie 9 ifjB U 1 Declassified and Approved for Release by NSA on '10-'1 '1- 20'1 2 pursuant to E O '135 26 vl DR Case # 54778 DOCID 4009733 TOP SECRB'f Published Monthly by PI Techniques and Standards for the Personnel of Operations JUNE-JULY 1976 VOL III Nos 6 and 7 WILLIAM LUTWINIAK PUBLISHER BOARD OF EDITORS Editor in Chief Arthur J Salemme 56425 Co llec t ion L I Cryptanalysis Language 1 89555 P L K8025 Emery W Tetrault 52365 Machine Support 332 S 1 Mathematics Reed Dawson 39575 Special Research Vera R Filby 71195 Traffic Analysis Frederic O Mason Jr 41425 For individual subscriptions send name and organizational designator to CRYPTOLOG PI TOP SECRET 86-36 TOP SECRET C r 'I tJlJTER l THE t9 EL1 T fELEf 1ETRY BUEJ ES L 86-36 Chief W ij w g 1 IF NSA ComputeJt and In60 lmaUon Suenc e6 lru -tUute in th Fuedman AUditOUUIn on 22 JanuMlj 1976 In order for you to calibrate what I have to say I want to make it very clear that I am not a computer expert I have never participated in the design of any computers I have had little operational experience with the care and feeding of them and I have never written software But I have had the opportunity to help design and work with a number of systems that use computers in very important ways and I am in a position now to affect the market for computers in the ELINT and telemetry business I'll be directing my remarks at uses not computers themselves When you don't intend to say anything very profound or may be afraid that you won't although it is all that you know you often call your talk or book an Introduction to I've resisted that today just to lure more of you here But it was suggested to me that a brief introduction to ELINT and telemetry might be in order I'll start with that and keep it vepy basic Then I'll phase into the description of a current set of computer applications and end with what I see as real trends in this area June-JulY 76 CRYPTOLOG Page 1 TOP SECRET EO 1 4 c P L 86-36 1l8bE VIA EURQPlIIlT EUR1 INlib Q lbJPY DOCID 4009733 fer SECRET June-July 76 CRYPTOLOG Page 2 TOP SECRET EO 1 4 c 1Jl tI4E1LE VIA' E6 -Aff38IkHHEL 6NLI -D OC I D 400t-'G9t-91r- i S 3 - L ---J -- --_ ---- --- ---- - - -- ---- ---- ------ --_ 1 4 c TOP SECRET 86-36 June-July 76 CRYPTOLOG Page 3 TOP E GRE T IiAliBLE vIA eeMIl4'f ellANNEL ' eULY DOCID 4009733 TOP SECRET June-July 76 CRYPTOLOG Page 4 Tep SECRET lIM SbE VIA E91 f lzt- n Jf 9NbY P L 86-36 DOCID 4009733 TOP SECRET EO 1 4 c P L 86-36 June-July 76 CRYPTOLOG Page 5 Tap SECRET ----------- IINIQblO Vl O bQJlHJT bW NNHS ElllbY DOCID L 4 4009733 TOP SECRET I c 86-36 June-July 76 CRYPTOLOG Page 6 TOP GBP clin J jON 1 8 IP l5QHINT l5IIANNEbS 81lbY - - - - - - - - ' _ - ' - - L _ _L -_ _ L _ _ DOCID 4009733 EO 1 4 c F L 86-36 P L 86-36 RIGHT ON VERA VVta Filby' -6 aJIilc1 e onJr acii o dec ep ti on How Vo We Know It'-6 TJtue CRYPTOLOG Feb uaAY 1976 ha-6 elicited muc h c omment h e cU the Agenc y and ha-6 even pMmpted a c oupie 06 people to put the fJr thought-6 on pap -6ee On Bung TJtuth6u t by -- ' JIC IRYPTOLOG ApJUt 1976 and Some PJUnupiu 06 Cov a nd Vec ep ti on by fCRYPTOLOG May 1976 The 6o t towing mU-6age 6 om G any -incii c cUu thcU the -6ubjec t 06 Mcii o dec ep ti o n -i -6 -i tdeed M M-6 o Filby po-inted out one 06 v-i ta t c onc t to oth Agenuu M we t t M OM own MGEN Sm-ith ha-6 gMnted h-i -6 p -6-ion to pubwh h-i -6 emMIv -in 6u t t Ed I'--_---l FM NCEUR TO NSOC EI2 DDF A A7 AS EO 1 4 c P L 86-36 FOR VERA FILBY FROMI'- l SUBJ HOW DO WE KNOW IT'S TRUE REF P L B CRYPTOLOG MAGAZINE FEBRUARY 197b PAGE b 1 VERA YOUR ITEM IN THE FEBRUARY 7b CRYPTOLOG RECEIVED MUCH INTEREST HERE AT STUTTGART BY COINCIDENCE UNLESS YOU PLANNED IT THAT WAY THE TIMING WAS EXCELLENT THE DIRECTOR OF INTELLIGENCE FOR USEUCOM MGEN H iP SMITH IS VERY INTERESTED IN SIGINT DECEPTION RECENTLY HE SENT A MESSAGE Tol J REQUESTING THAT THEY SEND A TEAM TO EUROPE TO BRIEF ON THIS SUBJECT SEE RE ALFA I SENT A COPY OF YOUR ARTICLE TO MGENSMITH AND ON 2't MARCH HE WROTE THE FOLLOWg 1 0 Et4H QUOTE P L 86-36 1 VERY INTERESTING RIGHT ON 2 IT SOUNPS LIKE THE AUTHOR IS NOT AWAREQf I--SO 'feF EeRE'f II'c'eee I THOSE TWO ORGANIZATIONS SHOULD BE PUT IN TOUCH WITH ONE ANOTHER 3 I WOULD SUGGEST ALL THE CRYPTOLOGIC ORGANIZATIONS GET TOGETHER FIRST AND POOL ALL THEIR KNOWLEDGE AND SPECIFIC EXAMPLES THEN ENLARGE A SUBSEQUENT MEETING TO INCLUDE GENERAL INTEL TYPES PERSONS WITH ACTUAL EXPERIENCE IN COMBAT ZONES WHERE COMM DECEPTION HAS BEEN USED 'to THEN CONDUCT COURSES IN EACH THEATER PREPARE TOOLS AND GUIDES END QUOTE EQllFIQENTIAIo IIVEEQ June-July 76 CRYPTOLOG Page 7 TOP SECRET 1I1' NBbE 'iII' E8tI'EN'f ElbltNNEhS 81lloY 86-36 DOCID 4009733 CONFIBENTIAL I To be or not to be That is the question Whether 'tis nobler to transliterate or render in the original That's not quite how the quote goes but the problem seems to have been around at least as long as Hamlet I would also venture a guess that we here at NSA have collectively expended considerably more nervous energy on our problem than Hamlet did on his It seems that every 3 to 5 years the spectre of transliteration raises its ugly head anew and all the decisions we made in the last skirmish have to be rediscovered and restated yet another time Why this problem refuses to stay solved is difficult to determine Could I suggest however that it is because we have not yet solved it Having spent about 10 years 1960-1969 transcribin analvzinv and reoortinQ Russian rr Editor's two-cents' worth B fore letting the author get into the body of hIS article the editor feels that it might Ibe advisable to take an explanatory side trip ito make certain that everyone knows exactly Iwhat transliteration is and what it isn't IWell it isn't translation and it isn't transIc ipti n Let's take the following hypothetical sItuatIon You're Mike You're on a tour of the Soviet Union You'd like to meet some 'Russian girls but your guide watches your busload like a hawk Finally one day you and another guy Joe manage to exchange a few remarks with a cute-looking girl standing in line outside of Lenin's Mausoleum You don't speak Russian so Joe helps out Here are a few of the interchanges Girl lJ aM lUie nOIJ eJlYM Mike What did she say Joe 1taA lUie nOIJ eJlYA Mike No I don't mean what did she say I mean what does it mean Joe Itm13ans Gimme a kiss _ _----Ir I am aware -- painfully aware -- that discussions of transliteration tend to get very emotional and highly Girl aR Te6g 30ByT Joe She wants to know your name provincial In an attempt to lower the emotional content Girl MaHR 30BYT Taila of this article let me begin by offering a Joe She says her name is Tamara definition of transliteration and then explainGirl KaR Ball HpaBKTCR MOCRBa ing some concepts and properties of translitera Joe She wants to know how we like Moscow tion Just then the guide shows up andsh6 E90J' 4 c Transliteration as I will be using the term both back onto the b u s P L 86- 3 6 refers to transforming textual information from one alphabet to another The properties of a Mike See you Tamara - same time same good transliteration scheme are that the scheme place should retain as much information about the On the bus y6u write down in your handyoriginal as possible and that it should be phrases n tebook for a non linguist you have exeasily learned and used by the persons using cell Jlta uditoryacuity and retentivity the particular scheme In particular I will Diamon' yeah putza Louie be addressinGBP only Russian-to-EnGBPlish translitKahk tibyah zuhvoot eration I - Minyah zuhvoot Tamara Kakh vahm nraveetsa Muskva Joe looks over your shoulder and says JPYou're spell ing everything all wrong He crosses out what YOU wrote and puts down June-July 76 CRYPTOLOG Page 8 CQNFIBBJTIAL 11A 'IQJ 1i VIA EURQI4HIT 61 lNHS eJf4Lt L 86-36 ------ DOCID 4009733 ---- CONFIBENTIAL Day mne potseZuy Kak tebya zovut Menya zovut Tamara Kak vam nravitsya Moskva Meanwhile back at the Mausoleum Tamara has jotted down two names in her notebook MaHK 4 0 A lot happened during these brief interchanges linguistically if not romantically speaking Let's extrapolate everything into a representation of the entire Russian language left-hand side Tamara and the entire English language right-hand side Mike TAMARA Rw ian pelied MIKE English direction Interpretation and translation are concerned with what the utterance means Mike and the NSA nonlinguist don't care a rat's whisker about how the linguist derives the meaning or how hard it is to master those complicated morphologic and syntactic rules For example Joe didn't bother to tell Mike that he changed Tamara's How does Moscow please itself to you to How do you like Moscow or to explain that he translated the name of 'the Russian city from 'IMoskva to English Moscow As far as the nonlinguist is concerned it's as easy for the translator to translate as it is for the transcriber to transcribe So the nonlinguist feels that that should be the end of the problem What then are those other five arrows doing in the chart They do not deal with the meanings of the words but only with their representation in printed form woJtci6 n CYRILLIC letieM Engw h woJtcio n pelied LATIN letieM When Mike asked What did she say and Joe answered She said n aH KHe no u eJIyA arrow Joe #1 that didn't help Mike much It was still in Russian When this process occurs in the COMINT business that is listening to people talking in Russian then putting down on paper in Russian what they said it is called TRANSCRIPTION The process is concerned sure enough with what the speaker said his exact words as he spoke them but it doesn't help the nonlinguist analyst any He still doesn't know what it means Incidentally even though this step doesn't yield the English meaning the voice transcriber has a hard job to do and he certainly has to know what the person is talking about before he can transcribe it What Joe did after his first little joke was to INTERPRET for Mike and Tama ra arrow Joe #2 Interpretation is translation usually back and forth from one spoken language to another People in the COMINT business arely are involved in interpretation Instead they are usually involved in TRANS LATIO the transformation of text in one written language to another written language -- and usually NSA translators specialize in the into When Mike decided for example to record the Russian sentences he wanted to remember arrow Mike #3 he did something similar to TRANSCRIPTION Joe #1 But instead of COMINT-style transcription writing down on paper the utterance in the original language as spelled in the original alphabet he tried to record the Russian sentences in Latin letters on the basis of the Russian pronunciation This can be done scientifically but when it's done by amateurs like Mike it usually looks weird When Joe looked over Mike's shoulder and corrected the spelling arrow Joe #4 he was -- we're finally hitting paydirt -engaged in TRANSLITERATION This is what this article once we get to it is all about the spelling of Russian words in Latin letters on the basis of their original Cyrillic speZling The problem is Who's got the one true system Joe used a system of transliteration in which the Russian letter n is represented by y If he had used the NSA system he would have written Vaj mne po t6e1 uj non-NSAers say that all those j' s look funny When Tamara wrote the two names in her notebook arrow Tamara #5 she TRANSCRIBED the names into Cyrillic on the basis of their sound She wrote YanK and 4 0 She would have been wrong arrow Tamara #6 to transliterate them according to their Latin speZZing MHKe and oe that would make them pronounceable in Russian as Meek-yeah and Yah-yeah Ridiculous isn't it And yet there are Engl ish words in the Russian language which are spelled in Cyrillic and pronounced in Russian in such a ridiculous way If Mike were ever to read Tamara's notebook and laboriously transliterate the two names into Latin characters arrow Mike #7 he would obtain Mayk Majk and Dzho Would he recognize his own name or Joe's Or would June-July 76 CRYPTOLOG Page 9 CONFIBENTIAL Ibid ShE VIA E8I1IN'F Elb'rIlIiEbS SInN DOCID 4009733 CONFIBENTIAL he be completely unaware that he had come up against one of the translator's biggest problems -- the rendition of non-Russian personal names place names etc not by transliteration but by restoring them to their original Latin spellings not Nikson Taym magazine Reno but Nixon TIME magazine and Renault Transliteration as we will discuss it then is a very specific operation the transformation of Russian text in Cyrillic characters into the Latin alphabet according to a specific preferred scheme The transliteration process does not produce a translation but carries into the Latin representation all the grammatical information present in the Cyrillic text Hence the meaning content of the Latin-translit eration text is identical with the meaning content of the original Cyrillic text and both texts would yield the identical translation Several Russian-to-Latin transliteration schemes have been developed Table 1 shows only a few of them The list of schemes is not intended to be exhaustive but rather to show the flavor of the transliteration world The various schemes conform to a greater or lesser degree to the requirements of accuracy that is nonambiguity and ease of use Of particular interest is the various handling of the Cyrillic letters E e llt ii x q ill m 'h b 3 JO R 14 out of the 33 letters in the Russian alphabet Note that Table 1 BoaY'd on Geogr aphic Ncones Russian A B B r IJ InteY'rlaLibY'a- tionaZ Y'y of StanCondaI'ds gY'ess 1 O1'gn A A A B V G D B V G D 8 V G D Y E E Y 2 E ZH ZH Z Z I I En E JK 3 H j K I M H 0 II Y E E Z Z I J I K L M N K L M N 0 P 0 0 P P K L M N R R S R C S S T T T U T U U P Y X uq ill In 'b F F KH TS CH SH HCH F KH TS CH SH SHCH H Y H Y C S SC Y D o most sc emes use com 1nat1ons 0 at1n characters to represent certain Cyrillic characters that is the Cyrillic and Latin characters do not constitute a one-to-one mapping 3 E Xl H YU IU YA IA I E E JU JA l Modified system which omits the diacritics used in the preferred LC sY5t m 2t rEwhenp d db ' Cit 'ifeCf- 1 4 c wise initial or when precedepby vowM L 86-36 YE or Y as noted abovY keep their data in different translite ation schemes then they shaZZ be taskedtllproduce a joint report or else be cal Se of their own respective missions they s1aZZ find it absolutely necessary to merge two or more of the computer data files involved here June-July 76 CRYPTOLOG Page 10 CONFIBENTIAL Il IiBbE VIA ESlIHi'F Elh' 'NNEbS SIIJoY -- - - ' - _ L _-L ---c -' - - -- -- --- ' DOCID 4009733 -' - -_ _ _- EONFIBENTIAL 1 4 c L 86-36 June-July 76 CRYPTOLOG Page 11 COIIFIBEiHIAL IWlBj 'lID ESIIIN r EII I NlIEbS B lbY DOCID Ll 4 4009733 EONFIBENTIAL c 86-36 June-July 76 CRYPTOLOG Page 12 GOfJF Il3ENTIAL ao N8bE VIA ESPlUI'f EI PlIlHS SlibY DOCID _ _ _ I _ _ _--L 4009733 sur IBENTIAL COMMENTS AS REQUESTED MtI-re on the AG-22 IATS P L 86-36 At the end 06 h -iA aJL ti c i e MlL6ing 6 AbolLt the AG- 22 1ATS CRYPTOLOG MMch 1976 Cecil Phil lip 6 U k ed Comme n -U anuone We1 l commen t 6 he ' 6 Jteceived I What' 6 WJtong with I A G - 22 IA T S -a-pp-ea Jt ed n the May 1976 6Ue 06 CRYPTOLOG The 60Uowing com- merLt 6 w ch weJte Jtecen t1y Jteceived 6 eem to be woJt th publMhblg n 6uLt de 6pile theiJt 6tight oveJ 1a pp ng 6 n tJtea tment CRYPTOLOG would cO l1 Ue to welcome 6uJr theJt commen t 6 06 a 6u b 6ta ve n a tu Jte all 6u bject Ed June-JulY 76 CRYPTOLOG Page 13 etlNFlflENTI %o 1 4 dlMlShE VIA E81 IIIi'f ElljlfNNEhS 81lhY P L 86-36 DOCID 4009733 COiJFIBEfHIAL P L 86-36 EO 1 4 c June-July 76 CRYPTOLOG Page 14 CONFIBENTIAL IIMIBbE YlA E811I11f EIWINI b S QNbJPY DOCID 4009733 GQ FIBENTIAL iii P L 86-36 EO 1 4 c June-July 76 CRYPTOLOG Page 15 cmWIBENTIAL jh NBIoE VIA E8 IIN'F EJhI iNEhS 8UIoY 1I 1 4 c DOCID 4009733 COUFIBENTIAL 86-36 June-July 76 CRYPTOLOG Page 16 COilFIBEfHIAL llMlBeE VIA CElIIIPlT CIlMlIlHS 614Li DOCIE I L 4 4009733 L c 86-36 mlFI BENT AL f j June-July 76 CRYPTOLOG Page 17 i t DOCID 4009733 UNCLASSIFIED NI'-CRII C NI f A J S The quotation on the next page was taken from a published work of an NSA-er The first letters of the WORDS spell out the author's name and the title of the work WORDS By DEFINITIONS A Allegedly the ones most likely to appreciate SST flights 2 wds 20 180 55 120 35 73 B Abnormally enlarged masses of lymphoid tissue at back of the pharynx 23 67 88 31 164 50 5 34 C Pseudonym of Emmanuel Poir French caricaturist 2 wds read as one word a Russian pun D --- man out game 95 170 174 E Covered stone bridge in Venice connecting ducal palace with the state prison 3 wds F What circling tigers in Little Black Sambo turned into 16 117 157 175 G Ignored 178 H French painter of flowers famous for paintings of roses 1759-1840 I Sleeping Woman dormant volcano in Mexico last eruption 1868 92 78 131 53 158 118 144 66 163 1 29 96 173 176 7 19 155 51 140 74 63 2 39 83 104 177 17 82 99 58 64 93 22 38 24 122 69 J State abbrv 46 150 K He opened up Japan 2 wds L Poem by Rudyard Kipling M Scout N Hot mixed drink topped with whipped cream 2 wds 70 132 O Popular U S dance for several decades prior to 1910 superseded by fox trot 40 145 159 171 43 149 8 60 P Despite what NSA linguist r s father Georgia late 1940' s said when son decided to 142 -3- 106 181 168 major in Spanish If English was good enough for Jesus it ought to be good enough for you the language spoken by Jesus 49 III Q Sharpen again R Japanese admiral at time of Pearl Harbor attack commander-in-chief of combined fleet S What showoff Fritz said to his brother they had been mountain-climbing with their mother 4 wds 75 136 166 108 11 T One of the mightiest rivers of the Indian subcontinent U Partially compacted granular snow that forms the surface part of the upper end of a glacier 153 147 30 116 June-july 76 CRYPTOLOG Page 18 UNClASSIFIED 86 119 48 _DOCID 4009733 UNCLASSIFIED V Where agreement that ended the War of 1812 was signed W Not discarded X Influence Y Highest peak in Wales Z Swiss tourist resort and munitions town something for everyone 161 52 89 169 Zl' Shooting at clay targets 70 N 71 E 72 K 73 A 74 I 75 S 6 7 B 68 T i lt 9 S K 42 53 t j 160 X 172 R 73 o oo o I o o Ii 61 Z 62 E 163 H 74 D lili J 164 J m@ r U 154 p 11 i l77 43 E 44 Y G 145 79 w 80 A 0 I'46 W' 55 I po Zl p7 F 158 G 1IIIIfI lllil I IIl u j 68 P jlllllllllllll lll ll 71 0 Y B 165 Y 66 S 167 E 75 F 176 I 177 N 178 G 78 G 79 K 80 C 6 7 1 81 P 69 Z 170 D 82 E 1I1 SoZution next month ---------- - _ _--------------------_ CMI BANQUET at GOVVARV SPACE FLIGHT CENTER F4iday 18 JUNE 1976 6 30 p m Ope n bAA -- p4ime 4ib cU nneJz --NASA toW TICKETS $8 50 CMI me mbeJz one gUe6t 06 CMI me mbeJz J $10 00 all Re6V vatioYIJ J Ro e CampbeU PI 3W090 x3957 June-July 76 CRYPTOLOG Page 19 UiKLASSIFIED oth J DOCIO 4009733 eeNfIJjENTIAl crHE c I1ARQUIS THE c I1EDIUM or Wow to core Code coverr 1 11111111'$ Mais monsieur Madame Claire was saying never have I been asked to dig so deep What is it that I must do She regarded my emerging wallet with suspicion The placards in the second-floor window had caught my eye Spirit communications vied with palmistry astrology tea-leaf readings and Egyptian sand divinations But for the on parle fran ais I'd not have entered tendering an ephemeral greenback Just one more Pierre-Simon madame I beg you Pierre-Simon de Fermat for just one little question I wanted to question Pierre-Simon de Laplace 1749-1827 the famous French mathematician' and astronomer about his notorious law of succession The law says that if an event has occurred k times in n trials at some fixed but unknown probability then the odds for the next trial are to be estimated as k 1 to n - k 1 in favor of occurrence Laplace had gone so far as to apply the law to the odds the sun would rise the morrow morn for which he had often been upbraided in limbo Try I urged And 10 after an agony of mutterings and eye rollings Madame Claire claimed to have the spirit of the Marquis de Laplace As a test I inquired politely about the rising sun It was a mistake An angry French voice issuing from the medium shot back Imbecile if you wish to consider collateral information then do' so I And wh n did this one die ' monsieur In 1665 Mon Dieu monsieur Adieu monsieur I rallied firing the twenty-dollar question Your law of succession doesn't seem to apply to many-celled multinomial distributions the kind where the number of cells is more than the sample size When one forms odds by adding unity to each cell count the odds on the unseen categories become impossibly large Man enfant he replied more in pity than anger naturellement one must count the live cells only Can you not estimate their number The medium woke with a gasp monsieur Au revoir C'est tout Mais c'est merveilleux Please one more and we will both become famous For further details see A Survey of Multinomial Estimation for Code Weighting NSA f TechnicaZ Juurnal Vol 20 No 3 Summer 197$ COHFIBEfffIAL June-July 76 CRYPTOLOG Page 20 eer fIJjErHIAL EO 1 4 c P L 86-36 DOCID 4009733 EO 1 4 c P L 86-36 Eb T SPOIEE -r l ' h T i l - - % - - - - - - - - - '''- - - - - - Y - ' - - - - - ------ oooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo Tim M u r ph y B 341 o ooo oooooooooooooooo ooo ooooooooo ooooooooooooooooooo o I June-July 76 CRYPTOLOG Page 21 SECRET SPOKE EO 1 4 c P L 86-36 I DOCID 4009733 SECRET SPO E EO 1 4 c F L 86-36 1 4 c L 86-36 June-July 76 CRYPTOLOG Page 22 SECRET SP81EE DOCID 4009733 SECRET SPOKE June-July 76 CRYPTOLOG SECRET SPSKE EO 1 4 c P L 86-36 Page 23 DOCID 4009733 E0 1 4 86 36 June July 76 Page 24 SECRET SPOKE DOCID 400973@ 1 4 c P L 86-36 HOW THINGS HAVE CHANGED P L I ll11 Bei di ial ea nun aoweil Baby i I I VI SU i i f VIRGINIA SUMS I ' 1 J As a member of the Cryptanalysis Career Panel I am interested in the selection and development of people with the best potential for becoming professional cryptanalysts In doing research in this field recently to trace the historical trends in cryptanalyst recruitment at the Agency I came across a description of one approach that had been taken by a competing agency not too long ago The description occurred in the second part of the two-part article On the Selection of Cryptanalysts by Alex Dettmann NSA Technical Journal Vol V No 1 January 1960 and No 2 April 1960 The author is identified as follows The author of this paper was a former lieutenant in the German Army As officer-incharge of the Russian Section of the Cipher Bureau of the German Army High Command he gave considerable thought to the selection and training of professional cryptanalytic personnel The paper which dealt with the historical development in the years 1935-1945 was written from memory after World War II Says Dettman Comments as requested - u v - Continued from page 17 6SMFIBEMTIAb Even with the danger of being considered old-fashioned and unobjective the author rejects the assignment of female personnel categorically and without exception as cryptanalysts though not as clerks or assistants Quite apart from the fact that the subject matter as such is foreign to a woman's mentality it must be added that it is extremely difficult for most women to engage in work about which no word may be spoken During his long years in a managerial status with the opportunity of becoming acquainted with most of the cryptanalytic positions in the former German Republic the author has never known a single woman who did even average work as a cryptanalyst All of the women who were assigned during and after 1943 as staff or intelligence assistants a bare total of about 100 can only be described as clerical personnel or more or less efficient assistants 11' '6 9 69Mr I BElH'IAb June-July 76 CRYPTOLOG Page 25 EO 1 4 c P L 86-36 SECRET SPOKE 86-36 DOCID 4009733 UNCLASSIFIED o Moo MM oo Mooooo MM oooooo AA oooo AA oooo AA oo o oo TT ooo TT oooo TT ooooooo HH ooo HHHHHHHHHHH o o M MM Mooooo MM ooooo I ' o o o AA ooo AA ooo AA oooooooo TT ooo TT oo TT oooooooo HH ooo HH ooooooo HH o MMMMMMMM oo MMMMMMMMM ooooo AA oo AA oo AA oo ooooooo TT oo TT oo TT ooooooooo HH oo HH oooooo HH o ooo MMMM ooo MM ooo MM ooo AAAAAAAAAAAAAAAAA ooo TTTTTTTTTTTTTTTTT HHHHHHHH HH ooo HHHH oo o o M MM Mooooo MM MM ooo AA ooooooooooooooo AA o TT ooooooooooooooo TT oooo HH ooo HH ooooooooo o Moo MM oo Moooo MM MM AA ooo AAAAAAAAAA oo A oo TT ooo TTTTTTTTTT oo T oooooo HHH oo HHHHHHHHHHH MMMMMMMMMM oooo MMM oooooooooooooo AAA ooooooooooooooooo TTT ooooooooo HHH HH H oooo HH o o MM oo MM ooooo MMM oooooooooooo AA oooooooooooooooooo TT oooooooooo HH HH ooo HH oo H oo HH oo oo MM oo MM ooooo MMMMM oo AAAAAAAAAAAAAAAAAAAA TTTTTTTTTTTTTTTTTTTT H HH ooo HH oo HHHH ooo oo o MMMM oooooo MM MM ooooooooooo AA ooooooooo TT oooo ooooo HH ooo HH ooo HHH ooo o o MM oo MM ooo MM oo MM ooooooo AAAAA ooooooooooooooo TTTTT ooooooooooo HHHH ooo HH oo HH HH oo MMM oo MMMM MM ooooo MM oooooooo AAA ooooooooo ooooo TTT ooooooooo ooo HHH ooo HH H oooo HHH I Like the twa-way IUllii o wwtwatc h mac une t UlnM atiorL MT a 6 language cortXinuv to 6a 6unate c eJz -taA n egmerL t6 06 the uertXi6ic - tec hrUc a1 and inte1-ligenc e c ommuniUv At iYLtVtv t in RUMian-to-Engwh MT tv eM we ob Vtve a new Mea 06 intVtv t -- Chinv e-toEngwh The 60UOW-trLg twa Mlic tu which deaf wdh the ame Chinv e-to-Engwh 1fT pJWjec t a J1 JUved at the eddolUaf 066ice in the ame week Wdh Agency intVtv t tike th -6 the edi to 6e c ompe1 ted to plUr bath aptides in thw e e ty dv pde the tigh t ov apping 06 the au tho ' commerL t6 Ed NOTES ON TRANSLATION FROM fH Ii CHINE - -S E_ _1R5f The Department of Computer Science of the United College of the Chinese University of Hong Kong has its own machine-translation project First announced offspring of the project is the CULT Chinese University Language Translator System which is being used we understand to translate inter alia Acta Mathematica Sinica In July 1975 we saw their translations of all six articles in the September 1974 issue We did not then or later see the original articles in their pristine ideographic form In January 1976 we found ourselves in licit possession of a nine-page package centered about a one-page article from the journal cited earlier A word perhaps two about the contents of those nine pages Page 1 was of course the original article characters and all and the characters in some cases were what the Chinese themselves call simplified in any case they're different from their traditional predecessors The next two pages contained the translation into English by some anonymous Chinese mathematician Following these were the two pages of input to the CULT System At least two more words are necessary here the computer never got to see the ideographs In their place it saw the appropriate fourdigit values from the well-known Standard Telegraphic Code STC Since the new simplified characters have not made their appearance in STC each one in the original journal article was replaced by the traditional one The computer never got to translate the title of the article the human's translation was entered in English along with the rest of the input All equations were copied presumably photocopied without any computer translation June-July 76 -cRYPTOLOG Page 26 UNCLASSIFIED P L 86-36 DOCIO 4009733 UNCLASSIFIED allowed to intervene Thus for example the character meaning or was replaced in one equation by or although that same character in STC guise occurred in the textual portion f the article and was indeed translated as or Back to the table of contents bei g The next two pages were the mildly edited Chinese original -- edited in the sense that the simplified characters have been replaced by their ancestors and certain shortcomings of the CULT System have been compensated in advance by the project workers As a tiny example the article begins by stating that certain alphabetic symbols our alphabet and also that of the Greeks stand for real numbers The Chinese are content to say a b c real numbers But in the edited input the simple string a b c was changed to a b and c Finally the last two pages contain the computer's translation of the mildly edited amplified historically adjusted original article In my opinion both translations -- its and his maybe hers but could the Chinese be that civilized -- were pretty good The fact that both translations in their own way were deficient allows me or at least gives me an excuse to belabor a vital point concerning technical translating i e translation of written material allegedly nonfictional which presupposes much relevant knowledge and relevant sophistication on the part of the ultimate human reader I consider my point vital enough to assign it its own paragraph The technical translator to be consistently correct and complete must be a master of at least three domains o the subject area be it a branch of mathematics or prison reform in the Third Circle o the source language particularly if not exclusively as it is used by native writers actively engaged professionally in the particular subject area and o the target larl Juage wi th the same particularity as that of the source language It might be objected that my point were it adopted as the criterion for choosing or retaining a technical translator would rule out all machine translation and many present human translators The objection is well taken Let me adduce a few examples from the translations of the Chinese article in question o The author describes certain geometric points by an adjective expressed by a single character which among its meanings and English translations has two nonsynonymous ones which are both valid technical ones in various mathematical areas In very specific area germane to the article at issue both the English terms show up but with slightly diverse meanings The character in the journal article was translated by the man by the correct English term for the specific context but the machine guessed incorrectly I emphasize that the machine guessed Even had the computer chosen critical rather than singular it would have had no reason for so choosing It was the man with his knowledge of the subject matter and the vocabulary precisely pertinent thereto who had peason to choose correctly o Another adjective this time with two characters fooled only this reviewer The adjective occurred in the title of the article which you recall was given to the computer as part of its input The use of the specific two characters jointly to form a single adjectival lexeme is only to be found to my knowledge in one dictionary published by the American Mathematical Society no less which gives the sole meaning of stationary No lexicographical source I could find gave the human translator's reading of autonomous Only by happenstance in translating a Russian article did I find a citation to the very current Americ m literature which for the exact situation at issue here uses autonomous My casual familiarity with the specific field was inadequate for purposes of technical translation o The human translator made a correct but infelicitous choice of words when he wrote of a limit cycle a geometric figure resembling a circle surrounding a particular geometric point More felicitous would be encircling The computer however was wpong on one of the two occasions and unnecessarily ambiguous on the other Of course some writers even Chinese are sometimes guilty of ambiguity or malfeasance Our present original writer was not Once the computer had a limi t cycle outside a point wpong and once the cycle contains the point Since the cycle at issue is itself made up of points to say that cycle Xcontains point Y may be saying that cycle X as a geometric curve encipcles point Y OP it may be asserting that cycle X as a collection of points boasts point Y as one of that collection To rephrase the problem there is no ambiguity when I say that point Y is on circle X or alternatively that point Y is within circle X but what is the situation when I say that Y is in X o A final example at least for the purposes of this overextended paper Here the computer dutifully and acceptably translated a phrase which the human without due cause omitted completely June-July 76 CRYPTOLOG Page 27 UNCLASSIFIED DOCID 4009733 UNCLASSIFIED The author had said that he could with no loss of generality standard procedure and verbiage for this sort of materi al set a particular symbol equal to a specific integer Compare The computer Without loss of generality we can take m 1 b ooooo- oooo oooo- The man we can t ake ill 1 The Anglophone mathematician prefers reading the assurance that generality is still secure UNCLASSIFIED ooooooooooooooooooooo o ooo- -ooo AN EVALUATION OF A SCIENTIFIC CHINESB MACHINE TRANSLATION I I Since the dawn of present-day computer science the idea that a computer should be able to translate one natural language into another has been a most fascinating problem to computer scientists and linguists alike A satisfactory solution to this problem however has proved to be most elusive Even when the concession is made that the first attempt at machine translation should be in scientific literature with its relatively small and precise vocabulary and usually uncomplicated syntax as opposed to the linguistically more complex artistic literature the solution is still far from complete Because of the importance of machine translation and perhaps also because of the enticing nature of unsolved problems many machine-translation projects exist throughout the world These projects are evaluated on the quality of the translated output as well as on the techniques employed in the translation algorithm The Agency with its high interest in language and language processing keeps abreast of developments in the area of machine translation and in line with this I was asked recently to evaluate the translations of a new Chinese-to-English machine-translation project The Chinese University of Hong Kong as a computer translation project CULT Chinese University Language Translator which is translating scientific Chinese into English This project has now been developed to the point that subscriptions are being accepted for their machine translations of two Chinese journals Acta Mathematica Sinica and Acta Physica Sinica Plans are being made to include translations of other scientific journals published in the People's Republic of China PRC in the CULT service In order to familiarize the world with this service and to demonstrate the quality of the product the organizers of CULT have prepared a paper describing their system the paper includes as an appendix an article from Acta Mathematica Sinica accompanied by both the CULT machine translation and for comparison a translation by a Chinese mathematician Also accompanying the machine translation is the form of the article at each step in the P 2 CULT process In an attempt to evaluate this machine-translation project I have compared the CULT translation with both the supplied human translation and my own translation The computer translation system employed by CULT involves only two steps before the English output is produced These steps are pre-editing and keypunching According to the CULT paper pre-editing involves the addition of some Chinese words such as understood subjects and predicates and the flagging of dependent Chinese clauses for the machine These additions are included so that the computer has a complete sentence to work with and so that the the extent of dependent clauses is known to the machine However in the sample article one English word was inserted into the Chinese text This occurred when the original article stated Let a l m n b be real numbers The preeditor inserted the English conjunction and into this sentence to yield the pre-edited version Let a l m n and b be real numbers While the insertion of English words at the preediting step is not mentioned in the CULT paper and its extent is unclear this example may only be the unavoidable result of slight differences in the form of enumeration statements in Chinese and in English One other interesting feature of the pre-editing step is the conversion of Chinese characters written in the shortened form now used by the PRC simplified characters back to their more traditional form The keypunching step is the conversion from this pre-edited form which is still in Chinese characters to a standard machine-recognizable form of Chinese This form the Standard Telegraphic Code STC merely assigns to each Chinese character a unique four-digit number The Acta Mathematica Sinica article chosen to demonstrate the capabilities of the CULT system states without proof a number of theorems in nonlinear differential equations While both the human translation and the machine translation included in the paper were quite good each made what I term a significant or a serious error A translation error is serious if it is unclear what the meaning of the original article June-July 76 CRYPTOLOG Page 28 UNCLASSIFIED P L 86-36 DOCID 4009733 CONFlfiENTIAL was based on the English translation alone This is quite different from the kind of error where the original meaning is apparent to a person familiar with the field The machine incorrectly translated one phrase as The system has a limit cycle outside two singular points whereas a correct wording would be something like The system has a limit cycle surrounding two singular points Realizing that a limit cycle is a kind of circle this error is definitely serious as the machine translation appears to make sense but is entirely misleading The human translator also committed a significant error when he rendered into English the phrase the sutonomous sic system Even allowing for a typographical mistake which yields the much more familiar word autonomous this is still a serious error as the original meaning was the nonlinear system While this is probably just a sloppy exchange of one technical term for another the effect on the English-speaking mathematican is just as serious as the effect of incorrectly rendered Chinese in the general mathematical conununity for a cover-to-cover translation of Acta Mathematica Sinica or any other of the PRC journals The answer to this question would have to be made by the would-be subscribers to the CULT translations UNCLASSIFIED _- ft _ LBTTBR TO TR EDITOR In view of the point I tried to make regarding NSA-ers' attempting to siphon personal mileage from the Vietnam tragedy Leo in October CRYPTOLOG JanUar 1976 I can only assume froml jarticle in April's CRYPTOLOG One Day in Danang Other Duties as Assigned that he still hasn't learned much in terms of the lessons of Vietnam His account of rockets smoke and shrapnel came across as a smokescreen not only for his lamentable lack of perception we covered that in the January issue but now also for his lack of good taste Too many silenced voicesw ill never be able to give such accounts not to anyone much less to a captive SIGINT readership One area where the human translator has an advantage over the computer is demonstrated by I chide both IforthisseC6rid P L the machine's translation of a phrase as a grossly indelicate effort and the publisher and limit cycle which contains the origin which editors of CRYPTOLOG for dignifying it as a would be more correctly translated as a limit feature article cycle which surrounds the origin It might be Respectfully said that this is a case of ambiRuouS translation Albert I Murphy Vl3 since the word contains in mathematics can E8liFIBEli'fIAh refer either to set containment as in 3' is contained in 2 3 t or to _geometric containment as in being contained inside of a circle I don't feel however that CULT should be expected to differentiate between Here's a Transcription in the News item these two distinct meanings or be criticized for from TIME magazine March 29 1976 p 24 not distinguishing them because the Chinese a very S l1g a v aIl4UY very shrewd original literally does say a limit cycle which The prosecutor also got help from a contains the origin The ambiguity IS In the highly unpredictable source William original Chinese not in the machine translaand Emily Harris Patty's most ubiquition The same Chinese words used here to mean tous S L A companions and the only members of the guerrilla band still alive geometric containment can sometimes mean set lew Times a pri ' containment The Chinese mathematician did use gave mag ill which Emily referr a the unambiguous word surrounds in this situa- sto relic in the shape of a monk tion but he possessed a knowledge of mathemat- f e that WiUie Wolfe once gave to Paty He called it an Olmec or someics and was able to use that knowledge in conthing she said olfe but long after his dealt junction with his translating ability The That offhand remark rang a bell c crying the tiny totem that h machine can base its decision only on the input with the prosecutor He recalled that on rently given her together with its stored dictionary which does one of the tapes that Patty made during Browning contended that that the her S L A sojourn she said looking up for the prosecu not include knowledge of the field of mathemat pigs probably had that Olmec mondway in the trial- some' ics It is therefore too much to expect the key that Wolfe-who died in the Los t een the time the judge allo machine to make such a differentiation What Angeles shootout on May 17 1974 traduce documents from h' -wore around his neck The fill tranyear the year before her arres is required or assumed is that the reader of that garbled remark as that had to take the Fifth so man' any article on nonlinear differential equations scribed acMonkey After reading t ew front of the jury should be able to resolve the ambiguity This TI ide Brow' the fBI What about the impact 0 on a hunc ally had a sim- dict on his future career I an is not in my opinion an unreasonable assumpilar Oimec iri iU ef in tier purse when I was 0ld6ysome people tliat tion I 11 86-36 had a fattn E ' E I O Having examined this sample article and the CULT translation of an entire issue of Acta Mathematica Sinica I feel that CULT could provide a quality translation of this and other PRC scientific journals on a timely basis The question remains however whether a need exists she was arrested The answer was yes That led to Browning's deft summation ploy when he wondered why Patty had testified that she could not stand '-- June-July 76 CRYPTOLOG Page 29 PI-Apr 76_53_24626 emw IBEfIT IAL wanted was a federal judi shouldn't try the case But I' that 1 could turn oVt r -' the trial to 1 UNCLASSI FlED P L 86-36 This document is from the holdings of The National Security Archive Suite 701 Gelman Library The George Washington University 2130 H Street NW Washington D C 20037 Phone 202 994-7000 Fax 202 994-7005 nsarchiv@gwu edu