THE RETRIEVAL PROBLEM IN PHOTOGRAPHY (1932)

by Emanuel Goldberg (1881-1970)

Translation and notes by Michael K. Buckland, School of Library and Information Studies, University of California, Berkeley, CA 94720-4600

Abstract: A new approach to mechanized document retrieval is demonstrated using optical and photoelectric methods to retrieve indexed documents stored on microfilm. Microfilm can be used to store images of documents compactly with suitable index codes recorded on the film adjacent to the image of the document. The film can be passed through a "statistical machine" [microfilm selector] in which a combination of light source, search template, and photoelectric cell is used to scan the moving film, identify the desired index code, project or copy the associated document. The machine demonstrated can search at up to 35,000 documents per hour, which could be increased to 100,000 per hour. [Translation of "Das Registrierproblem in der Photographie," a paper presented at the VIIIth International Congress of Photography, Dresden, 1931, and published in the proceedings in 1932.]

Translator's Introduction

This is a translation of Emanuel Goldberg's paper "Das Registrier Problem in der Photographie" presented at the VIIIth International Congress of Photography, Dresden, in 1931 (E. Goldberg, 1932b).

The paper presented here is of exceptional interest in relation to the development of information retrieval systems for four reasons:

1. It is a good, concise introduction to a form of information retrieval technology that was actively developed from the late 1920s to the 1960s: the use of photoelectric cells to search indexes on microfilm and to present a copy of the associated documents. Goldberg and other pioneers use the name "statistical machine," but later the name "microfilm selector" (or "microfilm rapid selector") was used.

2. Inasmuch as use of a photoelectric cell could be regarded as an electronic device, this paper is probably the first paper on electronic document retrieval and describes what was probably the first functional electronic document retrieval system, well before the development of electronic digital computers.

3. Priority in this area of information retrieval is generally assigned to Ralph R. Shaw or, more commonly, Vannevar Bush. This paper, along with his patents filed in 1927, clearly establishes Goldberg's priority, twenty years before Shaw and a decade before Bush.

4. This paper has been a "sleeper." It was presented in 1931 at an international congress, mentioned in some contemporary reports on the congress, and published in the proceedings. It has remained essentially unknown even though it was listed in a few of bibliographies and there are some passing references in the information retrieval literature to Goldberg having pioneered in this area. An English translation published in 1932 appears to have been even less well-known.

It was probably unhelpful that Goldberg used the German verb "Registrieren" (and other derivatives of "Register") which has various meanings including "to record" and, in relation to photography, "to align" as well "to index". The literal translation of the title of the English version of 1932--Methods of photographic registration--does not suggest a paper on document retrieval.

The new translation is rather literal in an attempt to reflect the flavor of the original. (Parentheses) are used as in the German original. {Braces} indicate the original German wording. [Brackets] mark the translator's additions. The advice of Traude Buckland and of Herbert Goldberg is gratefully acknowledged.

Emanuel Goldberg, born in Moscow in 1881, emigrated to Germany and became managing director of Zeiss Ikon AG in Dresden until forced by Nazis to flee first to Paris and then to Tel Aviv. He was responsible for a remarkable range of contributions to the theory of photographic processes, camera design, movie sound recording, extreme reduction "microdot" microfilm technology, and television (Browne & Partnow, 1983, 234-235; International, 1983, v. 2, p. 388; N. Goldberg, 1969; Neumann, 1957; Sipley, 1965, 58-59). He remained very interested in the use of photographic techniques for handling information (Neumann, 1957). In an era before digital computers he saw great potential in the approach to information retrieval described in the following paper, combining as it did the speed of light with storage capacity of microfilm.

The Retrieval Problem of Photography by Emanuel Goldberg

If photography is in itself a recording technique, a technique whose goal is to retain, to record natural appearances, so also there are particular kinds of photography in which the "recording" itself and not the image is the goal.

Soon after the discovery of photography it was recognized that the light-ray in conjunction with a photographic plate is an outstanding means to replace the moving indicator which is connected to a stylus and which draws a curve on paper. Gradually other special techniques were developed to serve the same purpose but they are typically photographic.

In particular, one should include among these techniques the so-called actinographic methods, in which the intensity of light is measured by greater or lesser darkening [Schwaerzung], which the effect of illumination produces on a light-sensitive coating. This actinographic process, which has, for example, been proposed for measuring the variation of natural light in the course of the year, has been very little adopted because the darkening is no direct measure of the light intensity. Rather, one must additionally measure the resultant darkening with the help of a light meter [Photometer] and establish it indirectly by way of the Characteristic Curve [Hurter and Driffield curve] of the light-sensitive coating.

This method can be greatly simplified if the light-sensitive film is not exposed directly to the light but is covered by a gray wedge-filter [i.e. a filter that is clear at one end and increasingly opaque towards the other]. Then the combined effect of the illumination with its fluctuating intensity and of the filter produces a spreading out of the darkening right on the developed plate such that a curve can be read directly. This method of intersecting {gekreuzten} filters has been frequently used, for example for the determination of the absorption curves of dyes, to measure the characteristic curves of photographic plates, meteorological studies, etc.

The new technology of low electric current has also opened new possibilities for recording using photographic means. Two tools are of quite special importance in this connection:

1. The photoelectric cell,

2. The cathode-ray oscillograph {Braunsche Roehre} [Braun tubes].

The latter provides the possibility of recording photographically those phenomena that happen in extraordinarily short periods of time. So, for example, bursts which take place in less than 1/1,000,000,000th of a second can be captured in this way. It seems also that sound recordings could also be achieved very simply with the help of the cathode-ray oscillograph.

In the photoelectric cell also an instrument is created which holds a host of currently unbounded possibilities for photographic recording. As one of these possibilities a new technique will be described here for the indexing [and retrieval] {zum Regustrierung} of any number of photographed documents.

There is an increasing trend to [micro]film the vast quantity of documents, checks, and messages which flow daily into offices and businesses before handling them any further. This work takes place with special equipment with which the individual items are transmitted one after another into the view of a special camera and then taken away for further handling. As film, normal 35 mm movie film is selected and the greatly reduced images can either be photographically enlarged or projected as needed.

With the great number of items that come to be involved the retrieval of an individual image is in practice very difficult because on a normal spool -- as used in movie equipment -- up to 50,000 images would be stored.

This retrieval can now be achieved by mechanical means by which when photographing each individual document a statistical [i.e. numerical] description is recorded with it, which contains the document number (tax number, check number, customer's account number) and, in addition, the day's date. This description can consist either of translucent letters, or numerals, on a black background (Fig. 1) or of suitably arranged [translucent] dots as, for example, in telegraph equipment where letters are represented by two or three holes in paper tape (Fig. 2).

[FIGS 1 and 2]

If a particular document is to be retrieved from the countless number photographed on the roll of film, then a positive [search] template [i.e. search card] {Schablone}, which bears the description of the desired document (Fig. 3) forms an image, with the help of a simple optical system, in the machine in the path of the film where the description (numerals or letters) is replicated (Fig 4).

[FIGS 3 and 4]

If the machine is now set in motion, the individual microfilmed images, each of which contain an image of the document and the associated description, run at a considerable speed (35,000 items per hour on the machine demonstrated at the Congress, easily raised to about 100,000 per hour) through the machine so that the [description of the] template is always projected on to the description of each document.

So long as the descriptions on the film differ from the characters on the template, the negative images on the descriptors do not completely cover the positive [translucent image of the] template. Only in one particular case, when the projection of the template is completely covered by the photographed descriptor, no projected light can pass through the film. If a photoelectric cell is mounted behind the moving film, it is continually struck by light and only at the point at which the template coincides with the microfilm descriptor does no light reach the photoelectric cell through the film. The photoelectric current is then interrupted for a brief moment and, indirectly through a relay, the moving film is abruptly stopped. The searched-for image, i.e. the document whose account number matches the inserted template, appears, greatly enlarged, on the projection screen.

The same result can be achieved in a another way where -- as indicated above -- in the place of numerals or letters a dot description is chosen. In a small field next to the document corresponding, for instance, to the soundtrack of sound film, several small clear circles are inserted, whose positioning corresponds to the identifying number. In this machine, instead of a template, a series of small electric light bulbs is envisaged, which, with the help of a dial, just like the dial of an automatic telephone, can be switched on in any selection (Fig. 5). One can, accordingly, simply through dialing a number, as when making a telephone connection, light up specific bulbs. These little bulbs, exactly as was the case in the previously described arrangement, form an image through a projection system in the position of the film, where the small translucent circles which correspond to the description of the document are located. As before, only in one particular case is the image of the lights covered by the translucent circles, and, of course, only if the bulbs lit by means of the dial coordinate exactly with the circles on the film. [Bulbs corresponding to the translucent circles are switched off and all others are switched on]. If a complete agreement of the unlit bulbs and the [translucent circles of the] film occurs, then the photoelectric current is here also interrupted and the document searched for is in thus found.

As described thus far, the arrangement has been demonstrated in action for the Congress. One can easily think of further development of the technique in which the document sought is not just projected but also reproduced immediately. Then, sitting by the telephone, one could dial a number, search, and copy any of hundreds of thousands of documents in a central archive which have something in common. The advantage of such an arrangement consists in that the original documents do not need to be disturbed again, that erroneous re-filing after use, as often occurs, is impossible, and that the documents can be arranged chronologically, without any real difficulties arising from it.

Translator's Postscript

This paper can be supplemented by Goldberg's patents (E. Goldberg, 1931; Zeiss Ikon & Goldberg, 1938). Two different prototypes were built but were not developed into a Zeiss Ikon product (H. Goldberg, 1990). I have not yet found any other material on Goldberg's statistical machine.

This paper was indexed, under "Photographie", in Internationale Bibliographie de Zeitschriftenliteratur (1932). It was included in two bibliographies published in the I.I.D. Communicationes of the International Institute for Documentation (Ergaenzung, 1935, 19; Schuermeyer & Loosjes, 1937, 25) and, in the U.S.A., in a bibliography in the Journal of Documentary Reproduction (Berthold, 1938, 100), but is otherwise remarkably absent from the literature of information retrieval. It is mentioned in the Goldberg Festschrift issue of the Bulletin of the Research Council of Israel vol 5C, no. 4, June 1957.

The English version (E. Goldberg, 1932a) is also listed by Berthold's bibliography (1938, 100), but seems to have remained even less well known.

For an attempt to reconstruct Goldberg's role in information retrieval see M. K. Buckland "Emanuel Goldberg, electronic document retrieval, and Vannevar Bush's Memex" elsewhere in this issue.

Translator's references:

Berthold, A. (1938). Selected bibliography on photographic methods of documentary reproduction. Journal of Documentary Reproduction, 1: 87-123.

Browne, T. & Partnow, E. (1983). Macmillan Biographical Encyclopedia of Photographic Artists and Innovators. New York: Macmillan.

"Ergaenzung und Verbesserung zu dem Artikel von Dipl. Ing. J. P. C. van Asperen ueber moderne photographische Reproduktionsverfahren." (1935) I.I.D. Communicationes 2(2): 16-19.

Goldberg, E. (1931). Statistical Machine. U.S. patent 1,838,389. Dec. 31, 1931.

Goldberg, E. (1932a). Methods of photographic registration. British Journal of Photography, 79: 533-534.

Goldberg, E. (1932b). Das Registrierproblem in der Photographie. In International Congress of Photography. 8th, Dresden, 1931. Bericht über den VIII. Internationalen Kongress fuer Wissenschaftliche und Angewandte Photographie, Dresden, 1931. Herausg. von J. Eggert und A. von Biehler. (317-320). Leipzig: Barth.

Goldberg, H. (1990). Personal communication.

Goldberg, N. (November 1969). The other Goldberg: A visit with Zeiss Ikon's practical genius. Popular Photography, 65, no. 5: 88-89 & 154.

International Biographical Dictionary of Central European Emigrés, 1933-1945. (1983). Munich: Saur.

Internationale Bibliographie der Zeitschriftenliteratur. Abt. A: Bibliographie der deutschen Literaturwissenschaft, 70 (1932).

Neumann, S. (1957). Prof. Emanuel Goldberg. Bulletin of the Research Council of Israel. Section C: Technology, 5C, no. 4: iii-v.

Schuermeyer, W., & Loosjes, T. P. (1937). "Literatur ueber die Anwendung von photographischen Reproduktionsverfahren in der Dokumentation." I.I.D. Communicationes 4(2): 23-29.

Sipley, L. W. (1965). Photography's great inventors. Philadelphia: American Museum of Photography.

Zeiss Ikon AG. & Goldberg, E. (1938). Vorrichtung zum Aussuchen statistischer und Buchhalterischer Angaben. [German] Patentschrift 670 190. Dec. 22, 1938.