Chapter V: Part 5

* Accept that, in a transition period, de facto standards will have
to be developed.

* Capture information in a way that keeps all options open and
provides for total convertibility: OCR, scanning of microfilm,
producing microfilm from scanned documents, etc.

* Work closely with the generators of information and the builders
of networks and databases to ensure that continuing accessibility is
a primary concern from the beginning.

* Piggyback on standards under development for the broad market, and
avoid library-specific standards; work with the vendors, in order to
take advantage of that which is being standardized for the rest of
the world.

* Concentrate efforts on managing permanence in the digital world,
rather than perfecting the longevity of a particular medium.

******

+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ DISCUSSION * Additional comments on TIFF * +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

During the brief discussion period that followed BATTIN's presentation, BARONAS explained that TIFF was not developed in collaboration with or under the auspices of AIIM. TIFF is a company product, not a standard, is owned by two corporations, and is always changing. BARONAS also observed that ANSI/AIIM MS53, a bi-level image file transfer format that allows unlike systems to exchange images, is compatible with TIFF as well as with DEC's architecture and IBM's MODCA/IOCA.

******

+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ HOOTON * Several questions to be considered in discussing text conversion * +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

HOOTON introduced the final topic, text conversion, by noting that it is becoming an increasingly important part of the imaging business. Many people now realize that it enhances their system to be able to have more and more character data as part of their imaging system. Re the issue of OCR versus rekeying, HOOTON posed several questions: How does one get text into computer-readable form? Does one use automated processes? Does one attempt to eliminate the use of operators where possible? Standards for accuracy, he said, are extremely important: it makes a major difference in cost and time whether one sets as a standard 98.5 percent acceptance or 99.5 percent. He mentioned outsourcing as a possibility for converting text. Finally, what one does with the image to prepare it for the recognition process is also important, he said, because such preparation changes how recognition is viewed, as well as facilitates recognition itself.

******

+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ LESK * Roles of participants in CORE * Data flow * The scanning process * The image interface * Results of experiments involving the use of electronic resources and traditional paper copies * Testing the issue of serendipity * Conclusions * +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

Michael LESK, executive director, Computer Science Research, Bell Communications Research, Inc. (Bellcore), discussed the Chemical Online Retrieval Experiment (CORE), a cooperative project involving Cornell University, OCLC, Bellcore, and the American Chemical Society (ACS).

LESK spoke on 1) how the scanning was performed, including the unusual feature of page segmentation, and 2) the use made of the text and the image in experiments.

Working with the chemistry journals (because ACS has been saving its typesetting tapes since the mid-1970s and thus has a significant back-run of the most important chemistry journals in the United States), CORE is attempting to create an automated chemical library. Approximately a quarter of the pages by square inch are made up of images of quasi-pictorial material; dealing with the graphic components of the pages is extremely important. LESK described the roles of participants in CORE: 1) ACS provides copyright permission, journals on paper, journals on microfilm, and some of the definitions of the files; 2) at Bellcore, LESK chiefly performs the data preparation, while Dennis Egan performs experiments on the users of chemical abstracts, and supplies the indexing and numerous magnetic tapes; 3) Cornell provides the site of the experiment; 4) OCLC develops retrieval software and other user interfaces. Various manufacturers and publishers have furnished other help.

Concerning data flow, Bellcore receives microfilm and paper from ACS; the microfilm is scanned by outside vendors, while the paper is scanned inhouse on an Improvision scanner, twenty pages per minute at 300 dpi, which provides sufficient quality for all practical uses. LESK would prefer to have more gray level, because one of the ACS journals prints on some colored pages, which creates a problem.

Bellcore performs all this scanning, creates a page-image file, and also selects from the pages the graphics, to mix with the text file (which is discussed later in the Workshop). The user is always searching the ASCII file, but she or he may see a display based on the ASCII or a display based on the images.

LESK illustrated how the program performs page analysis, and the image interface. (The user types several words, is presented with a list-- usually of the titles of articles contained in an issue--that derives from the ASCII, clicks on an icon and receives an image that mirrors an ACS page.) LESK also illustrated an alternative interface, based on text on the ASCII, the so-called SuperBook interface from Bellcore.

LESK next presented the results of an experiment conducted by Dennis Egan and involving thirty-six students at Cornell, one third of them undergraduate chemistry majors, one third senior undergraduate chemistry majors, and one third graduate chemistry students. A third of them received the paper journals, the traditional paper copies and chemical abstracts on paper. A third received image displays of the pictures of the pages, and a third received the text display with pop-up graphics.

The students were given several questions made up by some chemistry professors. The questions fell into five classes, ranging from very easy to very difficult, and included questions designed to simulate browsing as well as a traditional information retrieval-type task.

LESK furnished the following results. In the straightforward question search--the question being, what is the phosphorus oxygen bond distance and hydroxy phosphate?--the students were told that they could take fifteen minutes and, then, if they wished, give up. The students with paper took more than fifteen minutes on average, and yet most of them gave up. The students with either electronic format, text or image, received good scores in reasonable time, hardly ever had to give up, and usually found the right answer.

In the browsing study, the students were given a list of eight topics, told to imagine that an issue of the Journal of the American Chemical Society had just appeared on their desks, and were also told to flip through it and to find topics mentioned in the issue. The average scores were about the same. (The students were told to answer yes or no about whether or not particular topics appeared.) The errors, however, were quite different. The students with paper rarely said that something appeared when it had not. But they often failed to find something actually mentioned in the issue. The computer people found numerous things, but they also frequently said that a topic was mentioned when it was not. (The reason, of course, was that they were performing word searches. They were finding that words were mentioned and they were concluding that they had accomplished their task.)

This question also contained a trick to test the issue of serendipity. The students were given another list of eight topics and instructed, without taking a second look at the journal, to recall how many of this new list of eight topics were in this particular issue. This was an attempt to see if they performed better at remembering what they were not looking for. They all performed about the same, paper or electronics, about 62 percent accurate. In short, LESK said, people were not very good when it came to serendipity, but they were no worse at it with computers than they were with paper.

(LESK gave a parenthetical illustration of the learning curve of students who used SuperBook.)

The students using the electronic systems started off worse than the ones using print, but by the third of the three sessions in the series had caught up to print. As one might expect, electronics provide a much better means of finding what one wants to read; reading speeds, once the object of the search has been found, are about the same.

Almost none of the students could perform the hard task--the analogous transformation. (It would require the expertise of organic chemists to complete.) But an interesting result was that the students using the text search performed terribly, while those using the image system did best. That the text search system is driven by text offers the explanation. Everything is focused on the text; to see the pictures, one must press on an icon. Many students found the right article containing the answer to the question, but they did not click on the icon to bring up the right figure and see it. They did not know that they had found the right place, and thus got it wrong.

The short answer demonstrated by this experiment was that in the event one does not know what to read, one needs the electronic systems; the electronic systems hold no advantage at the moment if one knows what to read, but neither do they impose a penalty.

LESK concluded by commenting that, on one hand, the image system was easy to use. On the other hand, the text display system, which represented twenty man-years of work in programming and polishing, was not winning, because the text was not being read, just searched. The much easier system is highly competitive as well as remarkably effective for the actual chemists.

******

+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ ERWAY * Most challenging aspect of working on AM * Assumptions guiding AM's approach * Testing different types of service bureaus * AM's requirement for 99.95 percent accuracy * Requirements for text-coding * Additional factors influencing AM's approach to coding * Results of AM's experience with rekeying * Other problems in dealing with service bureaus * Quality control the most time-consuming aspect of contracting out conversion * Long-term outlook uncertain * +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

To Ricky ERWAY, associate coordinator, American Memory, Library of Congress, the constant variety of conversion projects taking place simultaneously represented perhaps the most challenging aspect of working on AM. Thus, the challenge was not to find a solution for text conversion but a tool kit of solutions to apply to LC's varied collections that need to be converted. ERWAY limited her remarks to the process of converting text to machine-readable form, and the variety of LC's text collections, for example, bound volumes, microfilm, and handwritten manuscripts.

Two assumptions have guided AM's approach, ERWAY said: 1) A desire not to perform the conversion inhouse. Because of the variety of formats and types of texts, to capitalize the equipment and have the talents and skills to operate them at LC would be extremely expensive. Further, the natural inclination to upgrade to newer and better equipment each year made it reasonable for AM to focus on what it did best and seek external conversion services. Using service bureaus also allowed AM to have several types of operations take place at the same time. 2) AM was not a technology project, but an effort to improve access to library collections. Hence, whether text was converted using OCR or rekeying mattered little to AM. What mattered were cost and accuracy of results.

AM considered different types of service bureaus and selected three to perform several small tests in order to acquire a sense of the field. The sample collections with which they worked included handwritten correspondence, typewritten manuscripts from the 1940s, and eighteenth-century printed broadsides on microfilm. On none of these samples was OCR performed; they were all rekeyed. AM had several special requirements for the three service bureaus it had engaged. For instance, any errors in the original text were to be retained. Working from bound volumes or anything that could not be sheet-fed also constituted a factor eliminating companies that would have performed OCR.

AM requires 99.95 percent accuracy, which, though it sounds high, often means one or two errors per page. The initial batch of test samples contained several handwritten materials for which AM did not require text-coding. The results, ERWAY reported, were in all cases fairly comparable: for the most part, all three service bureaus achieved 99.95 percent accuracy. AM was satisfied with the work but surprised at the cost.

As AM began converting whole collections, it retained the requirement for 99.95 percent accuracy and added requirements for text-coding. AM needed to begin performing work more than three years ago before LC requirements for SGML applications had been established. Since AM's goal was simply to retain any of the intellectual content represented by the formatting of the document (which would be lost if one performed a straight ASCII conversion), AM used "SGML-like" codes. These codes resembled SGML tags but were used without the benefit of document-type definitions. AM found that many service bureaus were not yet SGML-proficient.

Additional factors influencing the approach AM took with respect to coding included: 1) the inability of any known microcomputer-based user-retrieval software to take advantage of SGML coding; and 2) the multiple inconsistencies in format of the older documents, which confirmed AM in its desire not to attempt to force the different formats to conform to a single document-type definition (DTD) and thus create the need for a separate DTD for each document.

The five text collections that AM has converted or is in the process of converting include a collection of eighteenth-century broadsides, a collection of pamphlets, two typescript document collections, and a collection of 150 books.

ERWAY next reviewed the results of AM's experience with rekeying, noting again that because the bulk of AM's materials are historical, the quality of the text often does not lend itself to OCR. While non-English speakers are less likely to guess or elaborate or correct typos in the original text, they are also less able to infer what we would; they also are nearly incapable of converting handwritten text. Another disadvantage of working with overseas keyers is that they are much less likely to telephone with questions, especially on the coding, with the result that they develop their own rules as they encounter new situations.

Government contracting procedures and time frames posed a major challenge to performing the conversion. Many service bureaus are not accustomed to retaining the image, even if they perform OCR. Thus, questions of image format and storage media were somewhat novel to many of them. ERWAY also remarked other problems in dealing with service bureaus, for example, their inability to perform text conversion from the kind of microfilm that LC uses for preservation purposes.

But quality control, in ERWAY's experience, was the most time-consuming aspect of contracting out conversion. AM has been attempting to perform a 10-percent quality review, looking at either every tenth document or every tenth page to make certain that the service bureaus are maintaining 99.95 percent accuracy. But even if they are complying with the requirement for accuracy, finding errors produces a desire to correct them and, in turn, to clean up the whole collection, which defeats the purpose to some extent. Even a double entry requires a character-by-character comparison to the original to meet the accuracy requirement. LC is not accustomed to publish imperfect texts, which makes attempting to deal with the industry standard an emotionally fraught issue for AM. As was mentioned in the previous day's discussion, going from 99.95 to 99.99 percent accuracy usually doubles costs and means a third keying or another complete run-through of the text.

Although AM has learned much from its experiences with various collections and various service bureaus, ERWAY concluded pessimistically that no breakthrough has been achieved. Incremental improvements have occurred in some of the OCR technology, some of the processes, and some of the standards acceptances, which, though they may lead to somewhat lower costs, do not offer much encouragement to many people who are anxiously awaiting the day that the entire contents of LC are available on-line.

******

+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ ZIDAR * Several answers to why one attempts to perform full-text conversion * Per page cost of performing OCR * Typical problems encountered during editing * Editing poor copy OCR vs. rekeying * +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

Judith ZIDAR, coordinator, National Agricultural Text Digitizing Program (NATDP), National Agricultural Library (NAL), offered several answers to the question of why one attempts to perform full-text conversion: 1) Text in an image can be read by a human but not by a computer, so of course it is not searchable and there is not much one can do with it. 2) Some material simply requires word-level access. For instance, the legal profession insists on full-text access to its material; with taxonomic or geographic material, which entails numerous names, one virtually requires word-level access. 3) Full text permits rapid browsing and searching, something that cannot be achieved in an image with today's technology. 4) Text stored as ASCII and delivered in ASCII is standardized and highly portable. 5) People just want full-text searching, even those who do not know how to do it. NAL, for the most part, is performing OCR at an actual cost per average-size page of approximately $7. NAL scans the page to create the electronic image and passes it through the OCR device.

ZIDAR next rehearsed several typical problems encountered during editing. Praising the celerity of her student workers, ZIDAR observed that editing requires approximately five to ten minutes per page, assuming that there are no large tables to audit. Confusion among the three characters I, 1, and l, constitutes perhaps the most common problem encountered. Zeroes and O's also are frequently confused. Double M's create a particular problem, even on clean pages. They are so wide in most fonts that they touch, and the system simply cannot tell where one letter ends and the other begins. Complex page formats occasionally fail to columnate properly, which entails rescanning as though one were working with a single column, entering the ASCII, and decolumnating for better searching. With proportionally spaced text, OCR can have difficulty discerning what is a space and what are merely spaces between letters, as opposed to spaces between words, and therefore will merge text or break up words where it should not.

ZIDAR said that it can often take longer to edit a poor-copy OCR than to key it from scratch. NAL has also experimented with partial editing of text, whereby project workers go into and clean up the format, removing stray characters but not running a spell-check. NAL corrects typos in the title and authors' names, which provides a foothold for searching and browsing. Even extremely poor-quality OCR (e.g., 60-percent accuracy) can still be searched, because numerous words are correct, while the important words are probably repeated often enough that they are likely to be found correct somewhere. Librarians, however, cannot tolerate this situation, though end users seem more willing to use this text for searching, provided that NAL indicates that it is unedited. ZIDAR concluded that rekeying of text may be the best route to take, in spite of numerous problems with quality control and cost.

******

+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ DISCUSSION * Modifying an image before performing OCR * NAL's costs per page *AM's costs per page and experience with Federal Prison Industries * Elements comprising NATDP's costs per page * OCR and structured markup * Distinction between the structure of a document and its representation when put on the screen or printed * +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

HOOTON prefaced the lengthy discussion that followed with several comments about modifying an image before one reaches the point of performing OCR. For example, in regard to an application containing a significant amount of redundant data, such as form-type data, numerous companies today are working on various kinds of form renewal, prior to going through a recognition process, by using dropout colors. Thus, acquiring access to form design or using electronic means are worth considering. HOOTON also noted that conversion usually makes or breaks one's imaging system. It is extremely important, extremely costly in terms of either capital investment or service, and determines the quality of the remainder of one's system, because it determines the character of the raw material used by the system.

Concerning the four projects undertaken by NAL, two inside and two performed by outside contractors, ZIDAR revealed that an in-house service bureau executed the first at a cost between $8 and $10 per page for everything, including building of the database. The project undertaken by the Consultative Group on International Agricultural Research (CGIAR) cost approximately $10 per page for the conversion, plus some expenses for the software and building of the database. The Acid Rain Project--a two-disk set produced by the University of Vermont, consisting of Canadian publications on acid rain--cost $6.70 per page for everything, including keying of the text, which was double keyed, scanning of the images, and building of the database. The in-house project offered considerable ease of convenience and greater control of the process. On the other hand, the service bureaus know their job and perform it expeditiously, because they have more people.

As a useful comparison, ERWAY revealed AM's costs as follows: $0.75 cents to $0.85 cents per thousand characters, with an average page containing 2,700 characters. Requirements for coding and imaging increase the costs. Thus, conversion of the text, including the coding, costs approximately $3 per page. (This figure does not include the imaging and database-building included in the NAL costs.) AM also enjoyed a happy experience with Federal Prison Industries, which precluded the necessity of going through the request-for-proposal process to award a contract, because it is another government agency. The prisoners performed AM's rekeying just as well as other service bureaus and proved handy as well. AM shipped them the books, which they would photocopy on a book-edge scanner. They would perform the markup on photocopies, return the books as soon as they were done with them, perform the keying, and return the material to AM on WORM disks.

ZIDAR detailed the elements that constitute the previously noted cost of approximately $7 per page. Most significant is the editing, correction of errors, and spell-checkings, which though they may sound easy to perform require, in fact, a great deal of time. Reformatting text also takes a while, but a significant amount of NAL's expenses are for equipment, which was extremely expensive when purchased because it was one of the few systems on the market. The costs of equipment are being amortized over five years but are still quite high, nearly $2,000 per month.

HOCKEY raised a general question concerning OCR and the amount of editing required (substantial in her experience) to generate the kind of structured markup necessary for manipulating the text on the computer or loading it into any retrieval system. She wondered if the speakers could extend the previous question about the cost-benefit of adding or exerting structured markup. ERWAY noted that several OCR systems retain italics, bolding, and other spatial formatting. While the material may not be in the format desired, these systems possess the ability to remove the original materials quickly from the hands of the people performing the conversion, as well as to retain that information so that users can work with it. HOCKEY rejoined that the current thinking on markup is that one should not say that something is italic or bold so much as why it is that way. To be sure, one needs to know that something was italicized, but how can one get from one to the other? One can map from the structure to the typographic representation.

FLEISCHHAUER suggested that, given the 100 million items the Library holds, it may not be possible for LC to do more than report that a thing was in italics as opposed to why it was italics, although that may be desirable in some contexts. Promising to talk a bit during the afternoon session about several experiments OCLC performed on automatic recognition of document elements, and which they hoped to extend, WEIBEL said that in fact one can recognize the major elements of a document with a fairly high degree of reliability, at least as good as OCR. STEVENS drew a useful distinction between standard, generalized markup (i.e., defining for a document-type definition the structure of the document), and what he termed a style sheet, which had to do with italics, bolding, and other forms of emphasis. Thus, two different components are at work, one being the structure of the document itself (its logic), and the other being its representation when it is put on the screen or printed.

******

SESSION V. APPROACHES TO PREPARING ELECTRONIC TEXTS

+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ HOCKEY * Text in ASCII and the representation of electronic text versus an image * The need to look at ways of using markup to assist retrieval * The need for an encoding format that will be reusable and multifunctional +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

Susan HOCKEY, director, Center for Electronic Texts in the Humanities (CETH), Rutgers and Princeton Universities, announced that one talk (WEIBEL's) was moved into this session from the morning and that David Packard was unable to attend. The session would attempt to focus more on what one can do with a text in ASCII and the representation of electronic text rather than just an image, what one can do with a computer that cannot be done with a book or an image. It would be argued that one can do much more than just read a text, and from that starting point one can use markup and methods of preparing the text to take full advantage of the capability of the computer. That would lead to a discussion of what the European Community calls REUSABILITY, what may better be termed DURABILITY, that is, how to prepare or make a text that will last a long time and that can be used for as many applications as possible, which would lead to issues of improving intellectual access.

HOCKEY urged the need to look at ways of using markup to facilitate retrieval, not just for referencing or to help locate an item that is retrieved, but also to put markup tags in a text to help retrieve the thing sought either with linguistic tagging or interpretation. HOCKEY also argued that little advancement had occurred in the software tools currently available for retrieving and searching text. She pressed the desideratum of going beyond Boolean searches and performing more sophisticated searching, which the insertion of more markup in the text would facilitate. Thinking about electronic texts as opposed to images means considering material that will never appear in print form, or print will not be its primary form, that is, material which only appears in electronic form. HOCKEY alluded to the history and the need for markup and tagging and electronic text, which was developed through the use of computers in the humanities; as MICHELSON had observed, Father Busa had started in 1949 to prepare the first-ever text on the computer.

HOCKEY remarked several large projects, particularly in Europe, for the compilation of dictionaries, language studies, and language analysis, in which people have built up archives of text and have begun to recognize the need for an encoding format that will be reusable and multifunctional, that can be used not just to print the text, which may be assumed to be a byproduct of what one wants to do, but to structure it inside the computer so that it can be searched, built into a Hypertext system, etc.

******

+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ WEIBEL * OCLC's approach to preparing electronic text: retroconversion, keying of texts, more automated ways of developing data * Project ADAPT and the CORE Project * Intelligent character recognition does not exist * Advantages of SGML * Data should be free of procedural markup; descriptive markup strongly advocated * OCLC's interface illustrated * Storage requirements and costs for putting a lot of information on line * +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

Stuart WEIBEL, senior research scientist, Online Computer Library Center, Inc. (OCLC), described OCLC's approach to preparing electronic text. He argued that the electronic world into which we are moving must accommodate not only the future but the past as well, and to some degree even the present. Thus, starting out at one end with retroconversion and keying of texts, one would like to move toward much more automated ways of developing data.

For example, Project ADAPT had to do with automatically converting document images into a structured document database with OCR text as indexing and also a little bit of automatic formatting and tagging of that text. The CORE project hosted by Cornell University, Bellcore, OCLC, the American Chemical Society, and Chemical Abstracts, constitutes WEIBEL's principal concern at the moment. This project is an example of converting text for which one already has a machine-readable version into a format more suitable for electronic delivery and database searching. (Since Michael LESK had previously described CORE, WEIBEL would say little concerning it.) Borrowing a chemical phrase, de novo synthesis, WEIBEL cited the Online Journal of Current Clinical Trials as an example of de novo electronic publishing, that is, a form in which the primary form of the information is electronic.

Project ADAPT, then, which OCLC completed a couple of years ago and in fact is about to resume, is a model in which one takes page images either in paper or microfilm and converts them automatically to a searchable electronic database, either on-line or local. The operating assumption is that accepting some blemishes in the data, especially for retroconversion of materials, will make it possible to accomplish more. Not enough money is available to support perfect conversion.

WEIBEL related several steps taken to perform image preprocessing (processing on the image before performing optical character recognition), as well as image postprocessing. He denied the existence of intelligent character recognition and asserted that what is wanted is page recognition, which is a long way off. OCLC has experimented with merging of multiple optical character recognition systems that will reduce errors from an unacceptable rate of 5 characters out of every l,000 to an unacceptable rate of 2 characters out of every l,000, but it is not good enough. It will never be perfect.

Concerning the CORE Project, WEIBEL observed that Bellcore is taking the topography files, extracting the page images, and converting those topography files to SGML markup. LESK hands that data off to OCLC, which builds that data into a Newton database, the same system that underlies the on-line system in virtually all of the reference products at OCLC. The long-term goal is to make the systems interoperable so that not just Bellcore's system and OCLC's system can access this data, but other systems can as well, and the key to that is the Z39.50 common command language and the full-text extension. Z39.50 is fine for MARC records, but is not enough to do it for full text (that is, make full texts interoperable).

WEIBEL next outlined the critical role of SGML for a variety of purposes, for example, as noted by HOCKEY, in the world of extremely large databases, using highly structured data to perform field searches. WEIBEL argued that by building the structure of the data in (i.e., the structure of the data originally on a printed page), it becomes easy to look at a journal article even if one cannot read the characters and know where the title or author is, or what the sections of that document would be. OCLC wants to make that structure explicit in the database, because it will be important for retrieval purposes.

The second big advantage of SGML is that it gives one the ability to build structure into the database that can be used for display purposes without contaminating the data with instructions about how to format things. The distinction lies between procedural markup, which tells one where to put dots on the page, and descriptive markup, which describes the elements of a document.

WEIBEL believes that there should be no procedural markup in the data at all, that the data should be completely unsullied by information about italics or boldness. That should be left up to the display device, whether that display device is a page printer or a screen display device. By keeping one's database free of that kind of contamination, one can make decisions down the road, for example, reorganize the data in ways that are not cramped by built-in notions of what should be italic and what should be bold. WEIBEL strongly advocated descriptive markup. As an example, he illustrated the index structure in the CORE data. With subsequent illustrated examples of markup, WEIBEL acknowledged the common complaint that SGML is hard to read in its native form, although markup decreases considerably once one gets into the body. Without the markup, however, one would not have the structure in the data. One can pass markup through a LaTeX processor and convert it relatively easily to a printed version of the document.

WEIBEL next illustrated an extremely cluttered screen dump of OCLC's system, in order to show as much as possible the inherent capability on the screen. (He noted parenthetically that he had become a supporter of X-Windows as a result of the progress of the CORE Project.) WEIBEL also illustrated the two major parts of the interface: l) a control box that allows one to generate lists of items, which resembles a small table of contents based on key words one wishes to search, and 2) a document viewer, which is a separate process in and of itself. He demonstrated how to follow links through the electronic database simply by selecting the appropriate button and bringing them up. He also noted problems that remain to be accommodated in the interface (e.g., as pointed out by LESK, what happens when users do not click on the icon for the figure).

Given the constraints of time, WEIBEL omitted a large number of ancillary items in order to say a few words concerning storage requirements and what will be required to put a lot of things on line. Since it is extremely expensive to reconvert all of this data, especially if it is just in paper form (and even if it is in electronic form in typesetting tapes), he advocated building journals electronically from the start. In that case, if one only has text graphics and indexing (which is all that one needs with de novo electronic publishing, because there is no need to go back and look at bit-maps of pages), one can get 10,000 journals of full text, or almost 6 million pages per year. These pages can be put in approximately 135 gigabytes of storage, which is not all that much, WEIBEL said. For twenty years, something less than three terabytes would be required. WEIBEL calculated the costs of storing this information as follows: If a gigabyte costs approximately $1,000, then a terabyte costs approximately $1 million to buy in terms of hardware. One also needs a building to put it in and a staff like OCLC to handle that information. So, to support a terabyte, multiply by five, which gives $5 million per year for a supported terabyte of data.

******

+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ DISCUSSION * Tapes saved by ACS are the typography files originally supporting publication of the journal * Cost of building tagged text into the database * +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

During the question-and-answer period that followed WEIBEL's presentation, these clarifications emerged. The tapes saved by the American Chemical Society are the typography files that originally supported the publication of the journal. Although they are not tagged in SGML, they are tagged in very fine detail. Every single sentence is marked, all the registry numbers, all the publications issues, dates, and volumes. No cost figures on tagging material on a per-megabyte basis were available. Because ACS's typesetting system runs from tagged text, there is no extra cost per article. It was unknown what it costs ACS to keyboard the tagged text rather than just keyboard the text in the cheapest process. In other words, since one intends to publish things and will need to build tagged text into a typography system in any case, if one does that in such a way that it can drive not only typography but an electronic system (which is what ACS intends to do--move to SGML publishing), the marginal cost is zero. The marginal cost represents the cost of building tagged text into the database, which is small.

******

+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ SPERBERG-McQUEEN * Distinction between texts and computers * Implications of recognizing that all representation is encoding * Dealing with complicated representations of text entails the need for a grammar of documents * Variety of forms of formal grammars * Text as a bit-mapped image does not represent a serious attempt to represent text in electronic form * SGML, the TEI, document-type declarations, and the reusability and longevity of data * TEI conformance explicitly allows extension or modification of the TEI tag set * Administrative background of the TEI * Several design goals for the TEI tag set * An absolutely fixed requirement of the TEI Guidelines * Challenges the TEI has attempted to face * Good texts not beyond economic feasibility * The issue of reproducibility or processability * The issue of mages as simulacra for the text redux * One's model of text determines what one's software can do with a text and has economic consequences * +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

Prior to speaking about SGML and markup, Michael SPERBERG-McQUEEN, editor, Text Encoding Initiative (TEI), University of Illinois-Chicago, first drew a distinction between texts and computers: Texts are abstract cultural and linguistic objects while computers are complicated physical devices, he said. Abstract objects cannot be placed inside physical devices; with computers one can only represent text and act upon those representations.

The recognition that all representation is encoding, SPERBERG-McQUEEN argued, leads to the recognition of two things: 1) The topic description for this session is slightly misleading, because there can be no discussion of pros and cons of text-coding unless what one means is pros and cons of working with text with computers. 2) No text can be represented in a computer without some sort of encoding; images are one way of encoding text, ASCII is another, SGML yet another. There is no encoding without some information loss, that is, there is no perfect reproduction of a text that allows one to do away with the original. Thus, the question becomes, What is the most useful representation of text for a serious work? This depends on what kind of serious work one is talking about.

The projects demonstrated the previous day all involved highly complex information and fairly complex manipulation of the textual material. In order to use that complicated information, one has to calculate it slowly or manually and store the result. It needs to be stored, therefore, as part of one's representation of the text. Thus, one needs to store the structure in the text. To deal with complicated representations of text, one needs somehow to control the complexity of the representation of a text; that means one needs a way of finding out whether a document and an electronic representation of a document is legal or not; and that means one needs a grammar of documents.

SPERBERG-McQUEEN discussed the variety of forms of formal grammars, implicit and explicit, as applied to text, and their capabilities. He argued that these grammars correspond to different models of text that different developers have. For example, one implicit model of the text is that there is no internal structure, but just one thing after another, a few characters and then perhaps a start-title command, and then a few more characters and an end-title command. SPERBERG-McQUEEN also distinguished several kinds of text that have a sort of hierarchical structure that is not very well defined, which, typically, corresponds to grammars that are not very well defined, as well as hierarchies that are very well defined (e.g., the Thesaurus Linguae Graecae) and extremely complicated things such as SGML, which handle strictly hierarchical data very nicely.

SPERBERG-McQUEEN conceded that one other model not illustrated on his two displays was the model of text as a bit-mapped image, an image of a page, and confessed to having been converted to a limited extent by the Workshop to the view that electronic images constitute a promising, probably superior alternative to microfilming. But he was not convinced that electronic images represent a serious attempt to represent text in electronic form. Many of their problems stem from the fact that they are not direct attempts to represent the text but attempts to represent the page, thus making them representations of representations.

In this situation of increasingly complicated textual information and the need to control that complexity in a useful way (which begs the question of the need for good textual grammars), one has the introduction of SGML. With SGML, one can develop specific document-type declarations for specific text types or, as with the TEI, attempts to generate general document-type declarations that can handle all sorts of text. The TEI is an attempt to develop formats for text representation that will ensure the kind of reusability and longevity of data discussed earlier. It offers a way to stay alive in the state of permanent technological revolution.

It has been a continuing challenge in the TEI to create document grammars that do some work in controlling the complexity of the textual object but also allowing one to represent the real text that one will find. Fundamental to the notion of the TEI is that TEI conformance allows one the ability to extend or modify the TEI tag set so that it fits the text that one is attempting to represent.

SPERBERG-McQUEEN next outlined the administrative background of the TEI. The TEI is an international project to develop and disseminate guidelines for the encoding and interchange of machine-readable text. It is sponsored by the Association for Computers in the Humanities, the Association for Computational Linguistics, and the Association for Literary and Linguistic Computing. Representatives of numerous other professional societies sit on its advisory board. The TEI has a number of affiliated projects that have provided assistance by testing drafts of the guidelines.

Among the design goals for the TEI tag set, the scheme first of all must meet the needs of research, because the TEI came out of the research community, which did not feel adequately served by existing tag sets. The tag set must be extensive as well as compatible with existing and emerging standards. In 1990, version 1.0 of the Guidelines was released (SPERBERG-McQUEEN illustrated their contents).

SPERBERG-McQUEEN noted that one problem besetting electronic text has been the lack of adequate internal or external documentation for many existing electronic texts. The TEI guidelines as currently formulated contain few fixed requirements, but one of them is this: There must always be a document header, an in-file SGML tag that provides 1) a bibliographic description of the electronic object one is talking about (that is, who included it, when, what for, and under which title); and 2) the copy text from which it was derived, if any. If there was no copy text or if the copy text is unknown, then one states as much. Version 2.0 of the Guidelines was scheduled to be completed in fall 1992 and a revised third version is to be presented to the TEI advisory board for its endorsement this coming winter. The TEI itself exists to provide a markup language, not a marked-up text.

Among the challenges the TEI has attempted to face is the need for a markup language that will work for existing projects, that is, handle the level of markup that people are using now to tag only chapter, section, and paragraph divisions and not much else. At the same time, such a language also will be able to scale up gracefully to handle the highly detailed markup which many people foresee as the future destination of much electronic text, and which is not the future destination but the present home of numerous electronic texts in specialized areas.

SPERBERG-McQUEEN dismissed the lowest-common-denominator approach as unable to support the kind of applications that draw people who have never been in the public library regularly before, and make them come back. He advocated more interesting text and more intelligent text. Asserting that it is not beyond economic feasibility to have good texts, SPERBERG-McQUEEN noted that the TEI Guidelines listing 200-odd tags contains tags that one is expected to enter every time the relevant textual feature occurs. It contains all the tags that people need now, and it is not expected that everyone will tag things in the same way.

The question of how people will tag the text is in large part a function of their reaction to what SPERBERG-McQUEEN termed the issue of reproducibility. What one needs to be able to reproduce are the things one wants to work with. Perhaps a more useful concept than that of reproducibility or recoverability is that of processability, that is, what can one get from an electronic text without reading it again in the original. He illustrated this contention with a page from Jan Comenius's bilingual Introduction to Latin.

SPERBERG-McQUEEN returned at length to the issue of images as simulacra for the text, in order to reiterate his belief that in the long run more than images of pages of particular editions of the text are needed, because just as second-generation photocopies and second-generation microfilm degenerate, so second-generation representations tend to degenerate, and one tends to overstress some relatively trivial aspects of the text such as its layout on the page, which is not always significant, despite what the text critics might say, and slight other pieces of information such as the very important lexical ties between the English and Latin versions of Comenius's bilingual text, for example. Moreover, in many crucial respects it is easy to fool oneself concerning what a scanned image of the text will accomplish. For example, in order to study the transmission of texts, information concerning the text carrier is necessary, which scanned images simply do not always handle. Further, even the high-quality materials being produced at Cornell use much of the information that one would need if studying those books as physical objects. It is a choice that has been made. It is an arguably justifiable choice, but one does not know what color those pen strokes in the margin are or whether there was a stain on the page, because it has been filtered out. One does not know whether there were rips in the page because they do not show up, and on a couple of the marginal marks one loses half of the mark because the pen is very light and the scanner failed to pick it up, and so what is clearly a checkmark in the margin of the original becomes a little scoop in the margin of the facsimile. Standard problems for facsimile editions, not new to electronics, but also true of light-lens photography, and are remarked here because it is important that we not fool ourselves that even if we produce a very nice image of this page with good contrast, we are not replacing the manuscript any more than microfilm has replaced the manuscript.

The TEI comes from the research community, where its first allegiance lies, but it is not just an academic exercise. It has relevance far beyond those who spend all of their time studying text, because one's model of text determines what one's software can do with a text. Good models lead to good software. Bad models lead to bad software. That has economic consequences, and it is these economic consequences that have led the European Community to help support the TEI, and that will lead, SPERBERG-McQUEEN hoped, some software vendors to realize that if they provide software with a better model of the text they can make a killing.

******

+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ DISCUSSION * Implications of different DTDs and tag sets * ODA versus SGML * +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

During the discussion that followed, several additional points were made. Neither AAP (i.e., Association of American Publishers) nor CALS (i.e., Computer-aided Acquisition and Logistics Support) has a document-type definition for ancient Greek drama, although the TEI will be able to handle that. Given this state of affairs and assuming that the technical-journal producers and the commercial vendors decide to use the other two types, then an institution like the Library of Congress, which might receive all of their publications, would have to be able to handle three different types of document definitions and tag sets and be able to distinguish among them.

Office Document Architecture (ODA) has some advantages that flow from its tight focus on office documents and clear directions for implementation. Much of the ODA standard is easier to read and clearer at first reading than the SGML standard, which is extremely general. What that means is that if one wants to use graphics in TIFF and ODA, one is stuck, because ODA defines graphics formats while TIFF does not, whereas SGML says the world is not waiting for this work group to create another graphics format. What is needed is an ability to use whatever graphics format one wants.

The TEI provides a socket that allows one to connect the SGML document to the graphics. The notation that the graphics are in is clearly a choice that one needs to make based on her or his environment, and that is one advantage. SGML is less megalomaniacal in attempting to define formats for all kinds of information, though more megalomaniacal in attempting to cover all sorts of documents. The other advantage is that the model of text represented by SGML is simply an order of magnitude richer and more flexible than the model of text offered by ODA. Both offer hierarchical structures, but SGML recognizes that the hierarchical model of the text that one is looking at may not have been in the minds of the designers, whereas ODA does not.

ODA is not really aiming for the kind of document that the TEI wants to encompass. The TEI can handle the kind of material ODA has, as well as a significantly broader range of material. ODA seems to be very much focused on office documents, which is what it started out being called-- office document architecture.

******

+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ CALALUCA * Text-encoding from a publisher's perspective * Responsibilities of a publisher * Reproduction of Migne's Latin series whole and complete with SGML tags based on perceived need and expected use * Particular decisions arising from the general decision to produce and publish PLD * +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

The final speaker in this session, Eric CALALUCA, vice president, Chadwyck-Healey, Inc., spoke from the perspective of a publisher re text-encoding, rather than as one qualified to discuss methods of encoding data, and observed that the presenters sitting in the room, whether they had chosen to or not, were acting as publishers: making choices, gathering data, gathering information, and making assessments. CALALUCA offered the hard-won conviction that in publishing very large text files (such as PLD), one cannot avoid making personal judgments of appropriateness and structure.

In CALALUCA's view, encoding decisions stem from prior judgments. Two notions have become axioms for him in the consideration of future sources for electronic publication: 1) electronic text publishing is as personal as any other kind of publishing, and questions of if and how to encode the data are simply a consequence of that prior decision; 2) all personal decisions are open to criticism, which is unavoidable.

CALALUCA rehearsed his role as a publisher or, better, as an intermediary between what is viewed as a sound idea and the people who would make use of it. Finding the specialist to advise in this process is the core of that function. The publisher must monitor and hug the fine line between giving users what they want and suggesting what they might need. One responsibility of a publisher is to represent the desires of scholars and research librarians as opposed to bullheadedly forcing them into areas they would not choose to enter.

CALALUCA likened the questions being raised today about data structure and standards to the decisions faced by the Abbe Migne himself during production of the Patrologia series in the mid-nineteenth century. Chadwyck-Healey's decision to reproduce Migne's Latin series whole and complete with SGML tags was also based upon a perceived need and an expected use. In the same way that Migne's work came to be far more than a simple handbook for clerics, PLD is already far more than a database for theologians. It is a bedrock source for the study of Western civilization, CALALUCA asserted.