Volume II Number 4, December 1995

Recording For The Blind And Dyslexic: The Development of An Internet Accessible Online Catalog

Steve Noble
Technical Consultant, Recording for the Blind and Dyslexic
Media Services Coordinator, University of Louisville Libraries

PROJECT BACKGROUND SUMMARY

Recording for the Blind and Dyslexic (formerly Recording for the Blind) has been providing audio recordings of educational texts since its founding in 1948. RFB&D's master tape library currently contains over 80,000 volumes, making it the largest resource of its kind in the world. Beginning in 1992, RFB&D initiated a project to develop an online public access catalog that would allow both institutional providers and individual borrowers to search its holdings and order recorded texts via the Internet. Also in 1993, a coordinated pilot project was begun to allow a limited test group of institutional sites and individuals the ability to use the catalog ordering mechanism in an effort to gather structured feedback on the usefulness of the system and suggestions for improvements. Although the project cannot be considered an overwhelming success, much insight has been gained as a result of our efforts and will be of considerable value in the development of a future improved version of RFB&D's online public access catalog.

RFB&D SERVICES

Recording for the Blind and Dyslexic is a not-for-profit service organization that provides educational and professional books in accessible media formats to people with print disabilities including blindness, low vision, mobility impairments, and perceptual learning disabilities such as dyslexia. RFB&D primarily provides texts in four-track audio cassette format, but also has an ever growing electronic text program which can supply a limited number of texts on computer diskettes. Works available in E-text format are usually books that have little value in audio form, such as dictionaries, reference works, and computer software manuals. All audio recordings are loaned to borrowers free of charge, while E-text books are sold to users for a nominal charge.

Most of the work needed to produce RFB&D's texts is supplied by trained volunteers who record books or do E-text editing at 30 different studios scattered across the United States. In 1994, over 4,400 volunteers contributed some 324,000 hours worth of volunteer services. The combination of a largely volunteer workforce and generous corporate and individual contributors explains how RFB&D was able to provide most of its services to 36,000 active borrowers free of charge.

BIBLIOGRAPHIC INFORMATION

Prior to 1991, the only publicly-available reference tool for finding recorded books in RFB&D's master tape library general public was a printed catalog. These printed catalogs could be purchased by schools, outside agencies, and private individuals, but naturally were of little help to most of our borrowers. Beginning in 1991, RFB&D began to produce a subscription service on audio cassette called the Quarterly Recorded Catalog, which offered information on the latest recorded and electronic texts in a non-print format for the first time. In the following year, RFB&D began to offer its Quarterly Recorded Catalog on computer diskette as well. RFB&D's holdings can now as well be found by using the National Library Service CD-ROM catalog at regional and branch NLS libraries.

Recording for the Blind and Dyslexic began cataloging their holdings on OCLC beginning in 1980, and have been using the current in-house database system called MINISIS since 1985. However, public access to RFB&D's bibliographic information by consumers via online environments became available only in this decade. Individual computer users can now access bibliographic databases containing information on RFB&D's holdings by three major routes:

(1) a paid subscription dial-up service provided by the American Printing House for the Blind called _CARL et al_;

(2) the Library of Congress online catalog known as LOCIS which contains the braille and audio publications listed through the NLS, now available on Internet at the LOCIS.LOC.GOV site, and;

(3) RFB&D's own Internet catalog at the WAIS.JVNC.NET site at port 4445.

NEEDS ASSESSMENT

In light of the fact that RFB&D's borrowers now have a number of avenues through which they may access our bibliographic information, one may question the need to design and maintain a catalog which is solely designed for user access. Although there may be several legitimate reasons for any library to produce and control its own public access catalog, the primary impetus behind the development of RFB&D's Internet catalog was to support an automated online ordering process. A second consideration was the need to create a user friendly interface for those not familiar with bibliographic search methods. Both of these concerns were supposed to be addressed in the Internet database.

Prior to the Internet system's start-up, orders were generally phoned in via a toll free service to RFB&D's Princeton, New Jersey Headquarters, or were printed out and sent by mail or FAX. The Internet catalog's ordering function was designed to provide more effective service in this regard, because borrowers would be able to search our catalog and order materials during the same session without having to make a phone call or printing out information. Additional benefits would include: no busy signals or being placed on hold while waiting for your call to be answered; borrowers on the West Coast or living in England or Australia would not have to worry about changes in time zones; users outside the U.S. who are not able to use the toll-free service would not have to make a long distance phone call, and; an automated ordering system would relieve some of the very heavy traffic now routed through RFB&D's phone system. Clearly, an integrated automatic ordering system connected to an Internet accessible catalog would go a long way toward the important goals of independent user access, quicker processing of consumer book orders, and less need for human intervention resulting in lower service costs and a more efficient use of limited resources.

SYSTEM DESIGN AND IMPLEMENTATION

In 1992, RFB&D hired the Princeton based Internet provider Global Enterprise Services (formerly JvNC) to design and start-up a library based catalog system. The principal designer of the system was Vikas Aggarwal, with additional assistance provided by David Wagner and Spencer Sun. The catalog system was designed as a Wide Area Information Server, or WAIS database, allowing full- text searching and operating under a client-server model using the Z39.50 protocol as an ANSI standard. The system was set up on a Sun Sparcstation-2 running SunOS (Sun's Operating System), with 64MB of RAM, 1GB of disk space, and having a 40-Mhz SPARC CPU rated at 28.5MIPS. The following specifications were used to guide system design:

1. The catalog system should be available and searchable on the Internet;

2. The user interface should support ASCII terminal types as opposed to bitmapped type graphic terminals;

3. The searching capabilities would feature form based user input, an "expert" command-line type of searching capability for more experienced users, and field based searching capability;

4. Responses to the query should be displayed in user selectable brief or full format;

5. An order form for electronically ordering a text would be available as part of the system.

The system was developed as specified, and was connected to the Internet via ethernet connection at WAIS.JVNC.NET port 4445, and was also made available through the WAIS.JVNC.NET gopher site under the Publishers Online directory.

Because the system design was a WAIS database, it did not have the inherent ability to perform Boolean searches or allow for truncation, unlike most library catalogs. Instead, the WAIS system is designed to search the database using an automatic word weighting algorithm with more specific or rare terms assigned a higher value. The search is done by finding the bibliographic records with the most word weights for each word in the query. The record has its score increased each time it contains a word in the query. The record best matching the search is given a score of 1000 with others getting proportionately lower scores. GES did attempt to introduce an updated WAIS version earlier this year that would have allowed Boolean searching, but other problems related to a new word weighting algorithm produced unpredicted errors and the earlier non-Boolean design had to be kept.

Although the typical WAIS system is based on full-text searching only, GES modified the WAIS software to allow for field based searching like most library catalog systems which allow for subject, author, and title searches.

BENEFITS OF THE GES CATALOG

One of the nice features of the GES design is that the form based field searching capability allows users who have no prior experience with online catalog systems to immediately begin making searches without having to spend time learning how to construct search statements. Once logged on to the system, the user gets a screen like this:

 [Search]                         [Help]                   [Quit] 
   Command-Line
    Full-Search
RFB Book Number
         Status
         Author
          Title
        Edition
      Publisher
           Date
           ISBN
          Level
          Dewey
        Subject
  General Notes
            ISN

With this kind of arrangement, all a novice user has to do is move the curser to the author, title, or subject field, type in some information, and then move the curser to the button and press enter. This is far more intuitive that highly structured catalogs like LOCIS where you will get no hits with a subject search like "find art history" just because you failed to enter it as "find art--history." Other benefits of this kind of search screen are the "full-search" field that will look for any term or group of terms throughout the entire bibliographic record, and the automatic scrolling "command-line" field that permits the user to enter the entire search statement on one line without having to move the curser to different fields. The user does not have to read through a manual to figure out basic search construction, nor does one have to memorize specialized searching codes or language syntax to find books.

Another benefit of the WAIS based weighting system is that a user retrieves a number of related texts in addition to the specific title or exact subject area entered. This allows the user to browse through a much larger list of items than that which would be retrieved on a typical Boolean based system, while at the same time having the items that specifically match the search at the top of the list. This kind of approach means that even a poorly constructed search will almost always come up with something relevant. The user can be fairly confident about finding the majority of texts on a given topic without having to reconstruct a search in different ways.

Of course, the major benefit of the GES catalog is the ordering function, which allows users to order books which they have selected by simply typing "O" while viewing the bibliographic record. This presents an order form upon which users enter their mailing information and borrower ID number. After completion of the form, the order is sent to RFB&D for processing. {Currently, the order function is reserved for members of the pilot test group who have been issued a password to call up the order form.}

PILOT FEEDBACK AND SYSTEM DRAWBACKS

Although the system came online in late 1992, it was considered a rough test version, and no announcement was made regarding its development. After a year of initial testing, a number of institutional users (mostly disabled student services offices on college campuses) were asked to be part of the initial test group using the catalog. Several blind and low vision users were also added to the group, and the official coordinated pilot testing began in early 1994.

In general, pilot feedback has been positive. There are, however, a number of nagging problems about the system that seem to be almost universal. Despite the previously mentioned benefits of the WAIS based system, many users who have already become accustomed to online catalog systems using Boolean searching methods find the WAIS word weighted searching method very confusing. Especially exasperating to many seasoned online catalog users is the fact that the more terms you enter, the broader the search becomes--the exact opposite of a Boolean system. And to top that, the system almost always retrieves more that 100 hits and then tells the user "please refine search," when in actuality there is no way to do this. Obviously this would aggravate any new user.

Another low point of the system is that it has next to nothing in the way of help screens. There are only two help screens to be exact, and they give only general help on keystroke and control sequence commands and the very basics of searching.

A few users have experienced problems with terminal emulation, but this is almost always remedied by using alternate control sequence patterns, such as using control-p to move your curser up one line if the up arrow does not work. The major problem with terminal emulation is that the system does not accept IBM3270 terminal emulation, so colleges that have only IBM mainframe connections cannot access the catalog.

The most serious drawbacks to the current system are not quite so apparent to the casual users, but are placing major stumbling blocks in front of any further improvements of the GES catalog. The root of many system problems is based in the inability of GES to maintain or update the catalog due to loss of key personnel that designed the original system. Even updating the database to include newly recorded texts--what should be a simple and regular procedure for any online catalog--is possible only with great effort. Additionally, although the WAIS database does generate an electronic order as a function of the catalog, the GES system was never able to be integrated into RFB&D's production system, and so does not meet one of the original criteria for development of the system. Currently, the electronic orders are sent to a mailbox at Princeton, and must be manually keyed into RFB&D's automated production system. It now appears that these kinds of basic design flaws cannot be corrected in the GES database, and that a new system will have to be designed. In the meantime, Recording for the Blind and Dyslexic will keep the present Internet catalog operational while it develops a new online catalog. Information gathered from this pilot program will be of great assistance in designing a new bibliographic database, and will help to insure that a vastly superior system becomes available in the near future.

DESIGNING A NEW ONLINE CATALOG

A number of important design features will be need to be considered in the future bibliographic system developed by Recording for the Blind and Dyslexic. Some of these are paramount to the creation of a truly effective and efficient system. Others may not be quite so important to the basic function of the catalog, but are valid concerns for constructing a user friendly database. Here are some considerations for the next catalog:

1. An accessible system on the Internet.

RFB&D remains committed to the concept of a user friendly system that works well with adaptive technology, is easily reachable via the Internet, and allows users to place orders while logged into the system. The catalog will most likely be accessible through a telnet connection, but an HTML based form for searching and ordering on the World Wide Web is also being considered;

2. A fully integrated system.

A vital consideration of a new design is that it must be integrated into RFB&D's automated production system. The ideal system will verify that individuals are registered borrowers before allowing them to place orders, and will automatically enter the order information into the production system without having to involve human operators. This will speed up delivery of texts, and free up expensive human time for other applications. Such an integrated system would also allow individuals to search for and order e-text materials and audio equipment using credit card numbers providing a secure system is generated. Additionally, an integrated system would allow users to mail comments and suggestions to RFB&D while using the catalog;

3. A bibliographic system.

Consumers like knowing as much information about books as possible before ordering them. This saves the user considerable time by being able to choose the most valuable books without having to spend many hours listening to the recorded text first. A system that merely gives the user a title, author, and a few subject headings is not really an acceptable alternative for that reason. A system that has the capacity to include summaries or abstracts is ideal;

4. A system owned and operated by RFB&D.

Many of the problems cited above with the current catalog could have been fixed if RFB&D had been able to control and maintain its own system. It is therefore imperative that any new system that is developed for RFB&D's users should be under the full control of its own personnel. This would expedite any revisions to the system and allow for regular updating of the bibliographic information;

5. System feedback to users.

One flaw of the current system is that consumers do not receive much in the way of system feedback to assure them that their order is being processed. A full service automated system would be able to verify ordering information for each user on an individual basis. An ideal system would allow borrowers to pull up their own personal account to see what texts they have ordered, and what texts they already have out;

6. User help and documentation.

For a system to be truly user friendly, it must be equipped with a good online help facility, with specific informative help screens for each operation explaining all commands and search constructions. Some form of retrievable documentation should also be made available at the host site. Additionally, a taped tutorial would be of great benefit to many users;

7. Saving results of database searches.

Many users prefer to download long lists of works on a related subject for later use. This allows them to sort through lengthy lists with a word processing program and saves considerable time in putting together a select list of valuable books to order. For this reason, a system that sends search results to users as e-mail, or allows them to retrieve it like a gopher or ftp text, is very useful;

8. The ability to narrow searches.

Although some users have grown to prefer WAIS based searching, most individuals would prefer some means of initiating a Boolean search which would limit items retrieved to a smaller number. Boolean searching will be a certain feature of the new system.

CONCLUSIONS

Despite the failure of RFB&D's current online Internet database to meet the needs of the organization and its users, the pilot program has been an important experiment, nonetheless. Much insight and valuable experience has been gained. Although the new system being developed may not be able to fully integrate all of the design considerations just mentioned, Recording for the Blind and Dyslexic plans to include as many of them as feasible. Current time tables are very tentative, but call for introduction of the new system during the 1997-98 fiscal year. In the interim, all interested parties are urged to continue using the old catalog if you find it useful. RFB&D welcomes any additional input you may have, so please let us know what features you believe are important to have incorporated into our future system.

Noble, S. (1995). Recording for the Blind and Dyslexic: The development of an internet accessible online catalog. Information Technology and Disabilities E-Journal, 2(4).