Using Sorted Lists for Error CheckingUniversity of Idaho Library Moscow, ID 83844-2350
In the course of preparing a bibliography, errors creep into the data. This paper describes the use of sorted lists for error checking. Often sorted lists are useful for spotting errors or inconsistencies in basic fields like the author and title fields. Sometimes special lists can be used to take advantage of patterns in the data.
Two bibliographic projects where this approach to error checking was used are described briefly. One project from the 1980s used mainframe technology; the other used PC technology. For the second project, "printing" lists to computer files greatly facilitated use of those lists. Although those lists were not HTML documents, they can be viewed with HTML browsers. HTML tags were added to provide formatting and highlighting. The dBase III+ program proved very well suited for the second project.
In the course of any bibliographic project, numerous errors creep into the data. Many times data is simply input incorrectly. Also, the sources for the data may be incorrect or inconsistent. Half the job is error checking, and a number of techniques must be used to correct the data. One technique that may prove useful is to study sorted lists of the author field, the title field, and other major fields. Sorted lists are very useful for spotting misspellings, inconsistent punctuation, inconsistent usage of abbreviations, transposition of numbers or letters, and other anomalies and inconsistencies. Here is aexample of a title list used in the course of preparing a bibliography of US Army manuals.
If a field is shown in a sorted list, large errors tend to appear at the top or bottom of the list and minor errors are much less obvious. For example, if the first word of a title began with a "3" in the place of an "E," that would be obvious in a sorted title list. Transposition of "ea" within a title would be less obvious.
Of course, all errors and inconsistencies are not obvious at once. By repeated reviewing lists and making corrections, more and more errors are caught and corrected. Also, as the person preparing the bibliography becomes more and more familiar with the data, errors and inconsistencies become more apparent.
Indexing University of Georgia Bulletins
In the early 1980s, I indexed bulletins of the University of Georgia and the Georgia State College of Agriculture. Those indexes are described inSidebar One. One purpose of the project was to try out the FAMULUS software package developed by the Pacific Southwest Forest and Range Experiment Station of the US Forest Service. FAMULUS was an innovative set of FORTRAN programs developed around 1970, which allowed individual users to create bibliographies using a mainframe computer. The user designed the basic record, created a file containing those records, and used the system to create a variety of reports.
Sorted lists were particularly useful for resolving two problems.
I had to pick up computer printouts at the computer center, which was distant from my home and my office. Picking up printouts was tedious and time consuming and disrupted my work. During the course of the project, I made scores of trips to the computer center.
Army Manuals Listed in BSIR
Recently, I prepared a list of US Army manuals, which were listed in theBibliography of Scientific and Industrial Reports (BSIR) during the late 1940s. The index produced is described inSidebar Two.
I used dBase III+ to prepare that index because a copy was available and because I am familiar with the dBase command language.1 The command language is a specialized programming language for manipulating dBase files. The features intended to facilitate printing reports can easily be used to create HTML-formatted documents. Sorted lists can be used to identify errors and inconsistencies in the master file. Rather than print lists, I used the dBase SET ALTERNATE option to "write" the lists into computer files on my PC. I viewed those files using a Web browser. With simple programs written in the dBase command language, I was able to sort the master dBase file and print fields in the records in the order I specified. I had the program add HTML tags to format and highlight the records.
Here are two examples showing thesame sorted title list created using the dBase command language.
Using Windows Effectively
The sorts of data checking and correction, which took days during the earlier project, took only seconds. I could run the web browser and dBase at the same time using Windows. It was easy to switch between those two programs. I reviewed a list "printed" to a computer file using a web browser. If I spotted an error, I simply toggled into the dBase program, make the correction, and toggled back to the web browser. Using computer files for working copies allowed me to work rapidly and effectively without having to break my concentration to get printouts from a printer.
Basic Sorted Lists
As with the earlier project, establishing correct titles was difficult. The army's changing conventions for listing titles and series titles on the covers and title pages of manuals lead to confusion. Also, the indexers who had prepared theBibliography of Scientific and Technical Reports were inconsistent and occasionally sloppy. Sometimes, they confused series names with titles and occasionally paraphrased titles. A list sorted by title helped me to distinguish between titles and series names and to spot cases where titles were paraphrased. A separate series name field was not necessary in this bibliography because the manual numbering system is coded to show the general subjects of the manuals.
Exploiting a Pattern in the Data
I assumed that PB numbers-the order numbers given to documents listed inBSIR-were assigned using a mechanical number-stamping machine as the documents were unboxed. I expected many of the PB numbers to be in sets of sequential numbers. I reasoned that:
I wrote a program that created a list in PB-number order showing the PB number, manual number, and part of the title on a single line. A blank line indicated a gap in the PB-number sequence, and a line printed in blue showed a duplicate PB number. Here isthat PB-number list. I checked that list against theNumerical Index to the Bibliography of Scientific and Industrial Reports (BSIR), v. 1-10, 1946 - 1948 which listed PB numbers with the volume and page numbers of the corresponding entry inBSIR.2 I found:
Using a Spelling Checker
The dBase program does not contain a spell checker. I wrote a simple program to write the manual numbers and titles from each citation to a plain-text computer file. Then I used the spell checker in Microsoft Word to check the titles. I toggled between the dBase master file and Word making corrections and then checking further. When I used the spell checker, I had already checked and rechecked all the entries. I was amazed at how many misspelled words the spell checker found. Of course, the spell checker also found many peculiar Army words and spellings which had been correctly transcribed.
Printing the Finished Bibliography
I wrote a dBase command language program to produce the finished bibliography in the form of a full-scale, finished HTML document and to "write" it to a file. The dBase TEXT...ENDTEXT command allows text which is to be printed verbatim to be included in the program. I simply incorporate the HTML to be included in the final document into the program using TEXT...ENDTEXT. That has made editing my bibliography simple.
During the course of a bibliographic project sorted lists can be very useful for error checking and standardizing entries. Printing those files to computer files and viewing them using a web browser can greatly speed and simplify the work. Also, using HTML allows you to highlight and format the entries for viewing.
1. Although dBase III+ is an antique, it is flexible and powerful program. It can easily be used to write full-fledged HTML-formatted documents. Since it is obsolete, thousands of licensed copies are available for the asking on campuses around the country. Also, many people are familiar with dBase programming.
2. Numerical Index to the Bibliography of Scientific and Industrial Reports, Volume 1-10, 1946-1948. (New York: Special Libraries Association, 1949). (Available from University Microfilm International, PB2-OP04032).
Bulletins and Circulars of the University of Georgia
Bulletins of the University of Georgia are a window on the history of the school. Over the years a great variety of publications were published in the bulletin series because they could be mailed cheaply. Thissample page from the index shows the diversity of material printed in the Bulletin.
The publications of the semiautonomous Georgia State College of Agriculture are very interesting because the history of that college is a classic study in academic empire building. The rapid expansion of the college and its programs and activities is clearly reflected in its publications.
I used the FAMULUS program to produce separate indexes to university Bulletin series and to the College of Agriculture publication series and a merged list:A FAMULUS Index to The Bulletin of the University of Georgia.(Athens, Georgia: Political Science Department, University of Georgia, 1984.)
A FAMULUS Index to Georgia State College of Agriculture Bulletins and Circulars. (Athens, Georgia: Political Science Department, University of Georgia, 1984.)
A FAMULUS Index to Bulletins and Circulars of the University of Georgia.
This project had a useful byproduct. The bulletins and circulars of the Georgia State College of Agriculture had been microfilmed by the Southeastern Land Grant College Library Microreproduction Project. However, publications of the college which had been issued in the University Bulletin before the College of Agriculture started its own publications series had been missed. With a small grant from the College of Agriculture, I organized those early agriculture publications and had them microfilmed as:University of Georgia Agricultural Extension Service Bulletin, Supplement.(Athens, Georgia: University of Georgia College of Agriculture, 1984.)The master copy of the film was sent to the National Agricultural Library.
Army Manuals listed inBSIR
Shortly after World War II a large variety of publications were released for distribution by the Office of Technical Services which was a precursor of the National Technical Information Service. They were listed in theBibliography of Scientific and Industrial Reports. Those reports are now available from the Photoduplication Service of the Library of Congress. However, they must be ordered using the PB number they were assigned by the Office of Technical Services.
I used the dBase III+ program to create a master database containing records describing U.S. Army manuals listed inBSIR.1 Then I used the dBase command language to write an HTML-formatted document for "publication" on the World Wide Web. A list in manual-number order is the key component of the bibliography since users probably will know a manual number and be looking for the PB number with which the manual could be ordered. Title and PB number listings are also provided. Color-coding is used in the manual-number listing to distinguish between technical manuals and technical bulletins which are interfiled. Color coding is used in the title and PB-number listings to distinguish between the various manual series.
The list ofArmy Manuals Listed in the Bibliography of Scientific and Technical Reports, 1946-1949is available on the web at:www.lib.uidaho.edu/bsir/doc_no.html