The various different search strategies described in this section are complementary - they are designed to be used together. If one doesn't work - try one of the others, or try it on a different resource. We use examples that are based on real search needs that we have encountered in our work on this project.
This is the most common search strategy and the most important as it can be applied to almost any online resource. It simply means that you specify the subject you are interested in by searching on a series of words. The words are those you would expect to find in an article title (or description) of your area of interest.
For example, say you are interested in the role of women in Buddhism. One approach is to divide this up into concepts and list the words that might be expected to be associated with these concepts.
| Concept | Women | Buddhism |
Keywords |
women woman female feminist |
Buddhism Buddhist Buddha India Tibet Thailand religion |
The point is that your area of interest may be defined by several words. If you restrict yourself to just women and Buddhism then you may miss an article/book called 'Experiences of Buddhist women' or 'A feminist critique of religion in Tibetan society'. Authors of articles and books can express the same ideas in several different ways.
For a real example, in the BIDS ISI Arts & Humanities database for 1993-6:
there are 239 articles with buddhism in the title, keyword or abstract there are 263 articles with buddhist in the title, keyword or abstract there are 106 articles with buddha in the title, keyword or abstract |
If you only used one of these keywords you would miss a large number of relevant articles. There is an easy way to overcome this problem which exploits the similarity of words describing similar concepts and it is known as truncation searching (or wildcards).
The words Buddhism, Buddhist and Buddha all describe closely related concepts and all share a common stem - Buddh. What we really want the search to do is to match on any word that starts with these letters. To do this we can use a wildcard character which can stand for any other letter or letters. The wildcard character can vary between systems but it is usually the asterisk, *. This allows us to truncate the search term like this:
buddh*
Returning to our BIDS search:
there are 607 articles with buddh* in the title, keyword or abstract |
Not only is this truncated search quicker but it is more reliable - if you select the right stem then you can search for things which are relevant but you hadn't even considered! For example, if you use the stem wom* to catch both women and woman you may also get:
womb, womanhood, womanish, womanize,
womankind, womanly, womenfolk, womenkind
Of course this can be bad thing if you accidentally incorporate other words (e.g. wombat) into your search hits. A dictionary can be very useful in choosing appropriate word stems for a search.
Although using truncation is an important stage in widening this search we also need to include other terms. Most systems allow you to specify several searches and then combine them using three operators: AND, OR and NOT
The AND operator allows you to find those items which are common to two different searches, for example:
women AND buddhism
will give you a list of articles that contain both women and buddhism
On BIDS this search returned several articles, including:
TI: THAI WOMEN IN BUDDHISM - KABILSINGH,C AU: MUECKE_MA NA: UNIV WASHINGTON,SEATTLE,WA,98195 JN: JOURNAL OF ASIAN STUDIES, 1993, Vol.52, No.1, pp.211-212 IS: 0021-9118 DT: Book Review |
Technically, the AND operator returns the intersection of the two search sets. One approach is to try to see the result as a Venn Diagram:
The AND operator returns only those articles in region A of the Venn Diagram. Each search is called a set and usually associated with a number. Different systems use different characters to represent the AND operator, BIDS uses +. Although you can also use &.
For example on BIDS, the women AND buddhism search is like this:
No. Last run Year range Hits Search 3 - 1993-1996 3 1 + 2 2 - 1993-1996 239 (buddhism)@TKA 1 - 1993-1996 5418 (women)@TKA |
Set 3 is produced by giving the set number 1, the operator + (AND) and the set number 2.
The OR operator is used to join together two search sets to produce a bigger search set. In our example, we want to join together the wom* search with another one including female and feminist (fem* should be fine) to produce the women concept set.
wom* OR fem*
On BIDS the OR operator is a comma ,
Then we can use an AND to combine this with the Buddhism concept (we will use Buddh* for simplicity for now). The result on BIDS is:
No. Last run Year range Hits Search 5 - 1993-1996 13 3 + 4 4 - 1993-1996 607 (buddh*)@TKA 3 - 1993-1996 10858 1,2 2 - 1993-1996 3324 (fem*)@TKA 1 - 1993-1996 8118 (wom*)@TKA |
Set 3 is the women concept creating by using an OR on sets 1 and 2. Set 4 is the Buddhism concept using a simple truncation search. Set 5 is where it all comes together.
Note that we got 13 hits, when we tried the simple women AND Buddhism we got just 3. In other words with the simple search we missed 77% of the things we found using the more complex search!
Sometimes a word, or truncated word stem, brings back lots of words we don't want. It is possible that using wom* could have brought back numerous articles on wombats. The final operator, NOT, allows us to get rid of those whilst keeping the ones we want.
To use NOT in this way we could say:
wom* NOT wombat
On BIDS the symbol for NOT is the minus sign, -
Trying this search in the BIDS BL Inside Information system we get:
No. Last run Year range Hits Search 3 - 1993-1996 20714 1-2 2 - 1993-1996 17 (wombat*)@TI 1 - 1993-1996 20731 (wom*)@TI |
We successfully removed the 17 occurrences of wombat*. In this case it wasn't terribly important but the principle can be useful and save a lot of time in some searches.
A few examples should illustrate the importance of choosing the right keywords. If you were searching for articles about the leader of Libya who would you search for:
Gaddafi, Qaddafi, Qudhafi, Kaddafi, Kadafi, Gaddaffi, Quaddafi, Gadafi, Quadafi, el-Qaddafi, Qadhafi
For the political history of Russia:
Russia, Soviet Union, Commonwealth of Independent States, CIS, USSR
Similarly:
Burma / Myanamar, Peking / Beijing
American spellings / phrases are a common problem:
organisation / organization, sulphur / sulfur, physiotherapy / physical therapy
On many systems there are words, known as stop words, which are automatically removed from your search expression. These are typically common words such as:
the, and, to, a, as, is, be
This can make it difficult to search for certain expressions that contain these words, e.g. "to be or not to be". However, there is usually a mechanism for forcing the system to look for these words - but you will probably have to read the system specific help information.
The most common form of person-based searching is an author search - where the keyword is the author's name. The main problems encountered are:
authors and database compilers are sometimes inconsistent in the way they refer to themselves. This most commonly occurs by omitting, or including, a middle initial. For example, a BIDS ISI Science Citation Index search for the author John Self returns completely different results depending on whether his middle initial is used.
A search for SELF_JA produces:
TI: WHAT IF THE COMPUTER DOESNT KNOW THE ANSWER AU: DILLENBOURG_P, SELF_JA JN: COMMUNICATIONS OF THE ACM, 1996, Vol.39, No.8, pp.103-105 TI: PEOPLE POWER - A HUMAN-COMPUTER COLLABORATIVE LEARNING-SYSTEM AU: DILLENBOURG_P, SELF_JA JN: LECTURE NOTES IN COMPUTER SCIENCE, 1992, Vol.608, pp.651-660: |
A search for SELF_J produces:
TI: TAGUS - A USER AND LEARNER MODELING WORKBENCH AU: PAIVA_A, SELF_J JN: USER MODELING AND USER-ADAPTED INTERACTION, 1994, Vol.4, No.3, pp.197-226 TI: THE ROLE OF STUDENT MODELS IN LEARNING ENVIRONMENTS AU: SELF_J JN: IEICE TRANSACTIONS ON INFORMATION AND SYSTEMS, 1994, Vol.E77D, No.1, pp.3-8 |
The two searches don't even have any common items! But it is the same person!
people change their names. A search for Micheline Beaulieu between 1993-6 in the BIDS ISI Social Science database produces 7 hits including:
TI: A NEW ERA FOR OPAC RESEARCH - INTRODUCTION TO SPECIAL TOPIC ISSUE ON CURRENT RESEARCH IN ONLINE PUBLIC-ACCESS SYSTEMS AU: BEAULIEU_M, BORGMAN_CL JN: JOURNAL OF THE AMERICAN SOCIETY FOR INFORMATION SCIENCE, 1996, Vol.47, No.7, pp.491-492 |
However, a search for Micheline Hancock-Beaulieu in the same database in the same year range produces 7 entirely different hits, including:
TI: INFORMATION-RETRIEVAL - NEW SYSTEMS AND CURRENT RESEARCH - LEON,R AU: HANCOCKBEAULIEU_M JN: JOURNAL OF LIBRARIANSHIP AND INFORMATION SCIENCE, 1995, Vol.27, No.3, pp.177-179 |
Note that on BIDS you should omit hyphens for double barrelled names: Micheline Hancock-Beaulieu is searched for as HANCOCKBEAULIEU_M
This strategy relies on the grouping together of people with similar interests into research (or other institutional) groups. The chances of another person sharing some of the same interests (and therefore writing about the same topics) as another person are much higher if they belong to a common group.
The most effective method to achieve institution-based searching is via the Web. The steps are:
This is a very straightforward strategy. If you find a journal which contains one or more articles or interest then it probably worthwhile having a quick look at other articles in the same journal or publication. This can be done very quickly by hand in the Library or via BIDS.
Alternatively you can use journals to narrow down a search in combination with another search term. Using stress as a keyword will have radically different results if you combine it with the Journal of Occupational Psychology compared with the Journal of Mechanical Engineering Science.
Citation searching is a powerful search strategy that is especially useful if either:
A citation is another term for a reference - when an author refers to a previous piece of work in the text of the article. They are usually gathered together at the end of the article in a long alphabetical list - although sometimes they are included as footnotes. Some database indices, such as BIDS ISI, include these citations - establishing links between articles. These links are not dependent on the articles having the same keywords - they are an indication that one author thought the cited article was in some way relevant.
What this means is that you can specify an article and then see a list of all the authors in the database who subsequently referred to that article. For example, asking for a citation search on BIDS ISI Science Citation Index for TWIDALE_M produces this list:
->TWIDALE_M, 1989, P 4 INT C AI ED AMST......................................1 TWIDALE_M, 1991, P INT C LEARNING SCI......................................1 TWIDALE_M, 1993, 3RD EUR C COMP SUPP.......................................1 TWIDALE_M, 1993, APR P NATO ARW SEG........................................1 TWIDALE_M, 1993, COMICLANCS23 U LANC.......................................1 TWIDALE_M, 1993, Vol.4, J ARTIF INTELL ED..................................1 TWIDALE_M, 1993, Vol.4, J ARTIFICIAL INTELLI...............................1 TWIDALE_M, 1993, P ECSCW93 MILAN...........................................1 TWIDALE_M, 1994, P CSCW 94 LOS ANGELE......................................1 |
The first thing to note is that the same warnings about inconsistent naming still apply, doing the same for TWIDALE_MB only produces 1 entry:
->TWIDALE_MB, 1995, Vol.3, ALT J ASS LEARNING T..............................1 |
Each entry refers to a different citation - although entries 6 and 7 are actually the same article - and the number on the right is the number of times that article has been cited. In general you would expect older articles to have been cited more frequently. Selecting one of these, the 1995 M.B. Twidale entry, leads us to the article:
->TI: COMPUTERS AND GEOSCIENCE COMMUNICATION
AU: WHALLEY_WB
NA: QUEENS UNIV BELFAST,SCH GEOSCI,BELFAST BT7 1NN,ANTRIM,NORTH IRELAND
JN: TERRA NOVA, 1995, Vol.7, No.6, pp.641-644
IS: 0954-4879
DT: Editorial
CR: BROWNING_P, 1995 Vol.2 p.34, ACTIVE LEARNING
BROWNING_P, 1995 Vol.7 p.383, TERRA NOVA
FORD_C, 1995, E COMMUNICATION 0926
HABESHAW_S, 1992, 53 INTERESTING THING
HARDISTY_J, 1993, COMPUTERISED ENV MOD
MCCLURE_M, 1995, E COMMUNICATION 0926
MISNER, 1991, SPREADSHEET PHYSICS
PITT_M, 1995 p.49, NEW SCI
TWIDALE_MB, 1995 Vol.3 p.75, ALT J ASS LEARNING T
WHALLEY_WB, 1995 Vol.2 p.25, ACTIVE LEARNING
|
This is the only article, so far, to have cited the 1995 article in the Journal of the Association for Learning Technology. Note that this journal is not even indexed by BIDS ISI.
Citation searching allows us to look forward in time from an article and see who is referring to that work. It is a useful way of finding out about new authors in a field.
People often repeat searches that others have done in the past. Usually there is no way to re-use these searches and so save time and effort. However there are cases when these searches are published so that we can successfully re-use them:
Whenever items have been organised into categories similar things are found next to each other. This is most obvious in a Library or Bookshop, where books on the same topic or by the same author are on the same shelf. In Lancaster Library the Classmark associated with each item allows you to look for things with the same, or a similar, Classmark. Hierarchical Indices on the Web, or a CD-ROM, also enable you to look for similar things. Some Web search engines also have a 'Show me similar Web pages' option.
Some types of articles are reviews of a particular field - summaries of the main trends and issues. These are particularly useful for quickly understanding an area or finding the classic references. On BIDS ISI you can restrict your searching to just reviews through the 'Limit searching by document type' restriction on the Options menu.
--------------------------------------------------------------------
-- Options Menu --
1 Alter year range/recent update search/display pausing
2 Switch to different subject database
3 Limit searching to specified language(s)
4 Limit searching by document type(s)
5 Information
|
This gives a large menu of different types of document. Selecting the Review option will then mean that your search will only find review articles. Combining this restriction with a normal keyword search should enable you to quickly find review articles about a particular topic (if they exist).
Some collections of references are published as books and pamphlets, these are also a useful means of taking advantage of other peoples' searches, for example:
Classmark: ZDB2 (Oversize Pamphlet) [Ask Enquiries] ISBN: 0948848014 MANN, Margaret Browsing : an annotated bibliography. Loughborough : Centre for Library and Information Management, 1986. - 41p. - (CLAIM reports, 53). Brief Author/Title: MANN, M/BROWSING:ANNOTATED BIBLIOGRAPHY 1 copy. |
There are several Web indices which have been set up to aid searchers by collecting, evaluating Web sites and other online sources. In the UK the best sources for these are collected at NISS (National Information Services and Systems) Web site in the Directory of networked resources.
Examples include:
These indices are the results of several people collecting and organising information. However, the Web also allows people to publish their own lists. It is quite common for a Web surfer to have a link on their Home page to their own list of bookmarks. These public bookmarks allow you to exploit their efforts at finding things. You can locate people's Home pages either by navigating to them or through a search engine.