Devising a Good Computer Search Request

Computerized searching is the interaction of a human with a computer to retrieve information stored in data bases. A scribbled note with the cryptic message “computer search lung cancer chemotherapy” can be interpreted in many ways. There are thousands of articles on this topic; does the researcher really want a list of all citations? More to the point, does he or she really want to pay for all of them? What type of cancer? What type of chemotherapy? Is there a specific antineoplastic agent? Is the researcher interested in experimental animal research or clinical human results, or both? Should one look for English-language articles only? How many years should be covered? What data base should be searched? Before searching a data base it is essential to develop a very specific search request. This is true whether you’re requesting an information specialist to perform the search or if you’re doing your own searching. The first step is to understand what data bases are. There are two major categories of data bases: bibliographic and factual. Bibliographic data bases, such as Medline, Biological Abstracts, Chemical Abstracts, and Science Citation Index, generate lists of citations, but not the documents themselves. Factual data bases allow the user to answer a question directly, supplying the information needed—for example, the amount of a chemical that is lethal to humans. A limited number of “full-text” data bases, containing books or journal articles in their entirety, have become available recently. The contents of a data base are controlled by the parameters set by the data base producer. For example, most computerized data bases do not access material written before the mid-1960s. It is impossible to retrieve the entire world’s literature on a given topic from one database or even a combination of data bases. In addition, a comprehensive search of multiple data bases will result in some duplication that usually can’t be eliminated. A computerized search saves time and allows for multiple concepts to be combined, but generally will not turn up information that can’t be found manually. A computer will not miraculously produce answers to every question. Data Bases Vary The information explosion has caused an astonishing growth in the number of data bases available. Some 400 data bases existed in 1979; today there are 3,400. Most of these data bases are produced independently by various commercial concerns, resulting in differences in cost, quality and style of indexing. The costs of online searching indude telecommunication fees, data base royalties, and citation charges, all of which vary greatly and can be quite high. For example, telecommunication fees are $8 to $11 per hour, but will increase in January 1988 if an access charge proposed by the Federal Communications Commission is approved. Data base charges range from $15 to $300 per hour, averaging about $64 an hour; online printing charges vary from 4 cents to $7 per citation, with an average cost of 34 cents per citation. Most users think these costs justified when they consider the alternative—a manual search through years of printed indexes. Manual searches are time consuming, but because most research. ers are familiar with the process, they may be disappointed with their first computer search. When you scan article titles listed under the most appropriate subject heading, you automatically reject titles that seem to be unrelated to the research topic. A computerized search is the result of a rapid sorting of the documents to find those with common character strings. After “reading” these terms and combining them in the manner instructed, the computer produces a list that can be displayed, printed or downloaded onto a disk. The computer does not think—at least not yet. Whatever the computer “reads” it interprets literally. “False drops,” information apparently unrelated to the topic, can occur for several reasons. Terms that seem to be unique to a single topic may, in fact, have different meanings or interpretations in another context. With data bases that use a thesaurus, indexing policies may affect what is retrieved. For example, staff of the National Library of Medicine assign an average of 9.1 subject headings to articles indexed for Index Medicus. These headings, which are used in searches, are divided into major and minor points. Searches done with terms that turn out to be of minor importance will retrieve material that does not seem to be applicable. The entire content of the article cannot be reflected in the title. Also, indexes and data bases often contain variations, mistakes and inconsistencies. Problems can occur if an appropriate term does not exist for a subject or if an article has been indexed using terms that another person might not use. As the subject expert, you should be familiar with the scope of the literature, how much is being written, and what vocabulary is being used. The same indexing terms usually will be used in both print indexes and their computerized counterparts, which have the added benefit of “key word” searching or the ability to access any word within a record’s searchable fields. Remember, the searcher is an information specialist, usually a professional librarian with a master’s degree in library science, information science, or both who has developed expertise in finding, information. The ability to manipulate online data bases skillfully results from training and experience. The searcher may not have an academic background in your area of research. Devising a Strategy What is the best strategy for formulating a search request? First, write a problem statement detailing exactly what you need and showing the relationships between the concepts you want searched. In our example, the request was for chemotherapy of lung cancer. It turned out that the scientist only wanted information on chemotherapy of lung adenocarcinoma using vincristine or cisplatin. While “chemotherapy—vincristine, cisplatin” would have indicated more precisely what was needed, it is still inadequate. Lists of terms that do not clearly indicate the relationship between the words can lead to misinterpretation, especially if a secretary or assistant is delivering the request. An example of a problem statement might be “compare the use of the antineoplastic agents vincristine or cisplatin in the treatment of lung adenocarcinoma in humans.” One of the advantages of computerized searching is that concepts can be combined easily to produce a tailored bibliography. The Boolean operators AND, OR and NOT establish relationships between terms or concepts: OR broadens searches by allowing the use of synonyms, while AND and NOT narrow search results. AND is used when two or more terms must occur together before the record is selected. NOT eliminates parts of the set. In our example, the searcher will use the operator OR for the various drug terms: vincristine OR leurocristin OR Oncovin OR cisplatin OR cis-Platinum to create a set of records in which any of these terms appears. To narrow the search to include only those documents with these drugs terms and the other specifics of the search request, the Boolean operator AND will be used lung neoplasms AND adenocarcinoma AND the set of drug terms. Include synonyms for each search term, and try to be as complete as possible. For example, yincristine is sometimes referred to as leurocristine or the brand name Oncovin. A good search request will include all three. You should also supply any relevant citations and key authors. The format of the information (books, journal articles, conference proceedings and so forth) will affect which data base to search and will determine if the question is even searchable. For example, if you need only journal articles on vincristine or cisplatin for adenocarcinoma, the MEDLINE data base may be sufficient. If you want government reports or conference proceedings, CANCERLIT may be useful. if you need data on animal experimentation rather than human data, BIOSIS might be appropriate. You may not be familiar with the range of data bases available and the contents of each, but a professional searcher will be. There are other ways to limit the results of the search. Many data bases are international, and include information in numerous languages; you can limit the search by specifying certain languages. Another common way to limit a search is by publication year. Be careful, however, in setting limits; too many will result in no “hits.” In our example, limiting the search to adult male non-smokers may be combining too many concepts at once, and valuable information may be missed during the search. Consider beforehand how much information you want. If possible, give a year as a cutoff date or indicate the approximate number of citations you want. It’s easy to underestimate the volume of pieces published on a topic; if you want only a few articles, give an actual figure (“a few” may mean 10 to one searcher and 50 to another). Think of search results as being on a linear scale with precision at one end and recall at the other. High recall will retrieve more citations, but will also produce more “chaff” along with the good citations. High precision will find citations that are relevant but will not retrieve articles that may be of marginal interest. In practice, however, search results will contain many more successful “hits” than “false drops” when there is clear and complete communication between the researcher and the searcher. Most searchers prefer to speak directly with the person who will be using the information to clarify, expand, or modify the request into an effective search strategy. Also at this time, the searcher can consult the thesaurus used for indexing to make sure the term you use is correct. The searcher may prefer you to be present when the search is run so you can immediately evaluate the results. There is one major caveat here—the meter is constantly running while you are online. Adding spontaneously generated search terms without first checking the thesaurus or developing a strategy could result in needless expense. if the search must be 4elegated to a secretary or an assisnt, you should realize that person becomes an intermediary speaking with another intermediary. Make sure that the assistant clearly understands the search process and the question being posed. Applying these steps to the search request in our example, we get the following: “The use of the antineoplastic agents vincristine or cisplatin in the treatment of lung adenocarcinoma in humans. Search the NLM MEDLINE database from 1985 to the present. Limit the search to English language articles or any articles having an English abatract. Vincristine is sometimes referred to as leurocristine or Oncovin. Cisplatin is often spelled cia-Platinum. The article by J. Carmichael in European Journal of Cancer and Clinical Onocology (1985) was on the topic.” In this search request, the specific information needed (the type of lung cancer, the drugs involved and their various names) is clearly stated, and limitations (human subjects, English language, data base and time frame) are given. By following the strategy we’ve outlined, you should achieve more satisfactory search results, whether you are doing your own online searching or requesting one from an information specialist.

Devising a Good Computer Search Request

Interested in reading more?

Become a Member of