REVIEWING SEARCH RESULTS & REVISING SEARCH STRATEGY

The process of reviewing search results is a very important part of the search process. Once you've completed an initial search from beginning to end--dividing the topic into concepts, finding various search terms for each concept, searching for each concept separately and then searching for records related to all concepts--you need to check the results to find out if your strategy was accurate enough. Just as any good writer makes revisions after writing a first draft, researchers should expect to revise their search strategy after reviewing their initial search results. Checking your search results closely can give you good ideas of how to make improvements in your strategy so that in a succeeding search you can retrieve a greater number of records that are more precisely on your topic.

An initial search in any database should follow a search worksheet as closely as possible. In addition, each line of search terms (including logical, proximity and truncation operators, etc.) and the number of records retrieved for that search should be precisely recorded. Some database programs may record this information so that it can later be printed or downloaded with the records you select, but many do not. It is important to have an exact record of each search so that you will be able to consider ways to revise your search strategy after reviewing your initial results. A "Search Results Worksheet", such as the one shown on the following page, can help you keep a record of your searches.

The number of records in the set retrieved for each search is commonly referred to as the number of "hits" for that search set. As explained previously, it is common to have a large number of hits for a preliminary set that represents a single concept of a multiple-concept research topic. This type of set is usually created by ORing various synonymous or related terms (e.g. "information services OR libraries OR information superhighway".) After creating separate sets for each concept and then narrowing the search to records related to all of the concepts (usually with ANDs or proximity operators), the final number of records retrieved should be evaluated.

If the final number of hits is so large that it will take too long to read all of the citations, this may suggest that either your search is not precise enough or that your topic is too broad. How many hits is too many? This depends on various factors. If you are doing an extensive research project, it is reasonable to expect to be more thorough and take the time to review a larger number of records than if you are doing a short paper. If you are doing an in-depth research project, it may be reasonable to review 100 or more citations. If you don't need to be very comprehensive, you might not want to look at more than perhaps 30 records. Money can also be a factor. If you are searching on an online service, such as Dialog, it may not be cost effective to spend online time reviewing many records.

If your initial search strategy (creating sets for each concept and then looking for records related to all of your concepts) ends up with zero or very few hits, this could indicate either your topic is too narrow or your search is missing some significant terms. Check the number of hits for the preliminary sets representing each concept to see if any of the sets have a particularly small number of hits. Remember that your final set will never include more hits than the smallest preliminary set. If one concept is too narrow or lacks significant search terms, this will limit the number of hits in the final set.

Looking at the number of hits is just an initial method of appraising your results. You can get a much better indication of the effectiveness of your search by actually viewing the retrieved records. If the final number of hits is relatively small, you can look through all of the records. If there are too many records to be able to adequately look through them all, you should still look through some of the records--as samples--to get an idea of whether your search was on target. Most periodical databases list records in reverse chronological order, so the most recent articles are displayed first. This gives you a chance to focus on the most recent information first.

DATABASE:

SET # (if given)	SEARCH TERMS (indicate any fields used and include logical proximity & truncation operators, etc.)	RECORDS (# of Hits)
.	.	.
.	.	.
.	.	.
.	.	.
.	.	.
.	.	.
.	.	.
.	.	.
.	.	.

Displaying Records

Most database programs allow you to display retrieved records in various display formats. Most commonly, there are at least two formats: a shortened format that usually just includes the basic citation information, and a full-record format that includes the subject headings (or descriptors) for the record. If the database includes abstracts or the full-text of documents, you can usually select additional display formats that allow you to view either of those parts of the records.

When you have a significant number of retrieved records that you would like to view as quickly as possible, you can choose to just look at the short citation format for each record. When you see records that look interesting, you can then view the longer formats for more information.

As you view the retrieved records, you will quickly begin to see whether your search was very precise. If a good percentage of the records are relevant to your topic, then the search is relatively precise. It is especially useful to check the subject terms listed for any records related to your topic. This is often an excellent source for finding additional terms that can help you improve your search strategy.

Imprecise Searches

If there are many retrieved records that are not related to your topic--commonly referred to as "false hits" or "false drops"--you should try to figure out what part of your search strategy caused those inappropriate records to be retrieved. A significant percentage of false hits indicates that a search is not precise.

There are many possible causes of imprecise searches. Sometimes, you may use terms that have other meanings than those that you plan for. If those terms retrieve records based on their alternative meanings, these will be false hits. This problem can often be solved by linking these terms to other terms (using ANDs, proximity operators or NOTs) that limit their alternative definitions. It is important to be careful that the linkage does not eliminate some desired records, especially when using NOTs. It usually preferable to retrieve some records that off the topic than to eliminate relevant records. If the database includes full-text records or relatively long abstracts, some records may be retrieved based on terms that, although they appear in the record, are not related to the main ideas of the document. Proximity operators or field searching are usually required in these cases.

Inadequate Recall

If the final number of retrieved records is significantly less than you expected--based on your background reading or information from professionals or other sources--your search strategy may have failed to retrieve many records in the database that are relevant to your topic. This is referred to by professional researchers as poor "recall." Poor recall is often caused by either missing important search terms or using too many or inappropriate limiting operators or procedures (ANDs, NOTs, proximity operators or field searching.)

Too Many Hits

If your search strategy retrieves many more records related to your topic than you can possibly even browse through, this may indicate that your topic is too broad and you need to limit your search strategy in some way. One common way to limit a search strategy is by date. For example, you could limit your search to just documents published within the last year or two. This is particularly appropriate for rapidly changing topics such as those related to science or technology or to contemporary issues. You can also limit a topic by adding an additional concept to focus the search on a more specific aspect of a topic.

| Home | Syllabus | Assignments | Text | Student Projects | Instructor |

These materials may be used for educational purposes if you inform and credit the author and cite the source as: LSCI 105 Online Research. All commercial rights are reserved. To contact the author, send comments or suggestions to: Eric Brenner at brenner@smcccd.cc.ca.us