Step 3: Relationship Construction

The next step in the thesaurus construction process is to create relationships between the terms. There are three relationship types in a thesaurus, equivalence, hierarchical, and associative. The equivalence relationship is based between preferred and non-preferred terms, or those terms which refer to the same notion (Aitchison et al., 2000). This relationship type is dealt with in the final term selection step of the thesaurus construction and, therefore, will be examined in more detail later in this report.

The hierarchical relationship uses superordinate terms and subordinate terms to create a relationship between a class and its members (Aitchison et al., 2000). This relationship type is reciprocal and is seen in a thesaurus as broader and narrower terms or BTs and NTs. Broader terms are used to show the progression up the classification scheme while narrower terms show the progression downwards (Cleveland & Cleveland, 2001). This allows the user to more easily navigate through the thesaurus and find the appropriate term needed for their search whether it is a broader or a narrower search term which is needed. Hierarchical relationships are important in thesaurus construction because it improves performance in the areas of recall and precision (Aitchison et al., 2000).

There are a few different types of hierarchical relationships. One is the generic relationship; this links together a specific class and its members (Aitchison et al., 2000). For example, in my thesaurus the term libraries is the class and the different types of libraries are the members. Therefore, libraries becomes the broader term while academic libraries and public libraries are narrower terms. Another type is the whole-part relationship in which the name of the part implies the name of the whole. This relationship type is normally associative but there are some occasions in which it is hierarchical (Aitchison et al., 2000). For example, fields and disciplines can be whole-part relationships. In my thesaurus, library and information science, and history are narrower terms of fields and disciplines. Another hierarchical relationship type is an instance relationship which connects general categories and proper names (Aitchison et al., 2000). An example of this would be the broader term classification schemes with the narrower term Library of Congress Classification.

When I started to create the hierarchical relationships for my thesaurus I began by going off of my facet analysis and dividing those broad sections into smaller divisions. To be hierarchical, the terms need to be located under the same fundamental categories (Aitchison et al., 2000). I then consulted other thesauri, such as the American Society for Information Science and Technology (ASIS&T) thesaurus, the Library and Information Science Abstracts (LISA) thesaurus, and the Library, Information Science and Technology Abstracts (LISTA) thesaurus to get an idea of the broader terms and concepts that they used in their thesauri. I found that there were a lot of different ways to organize the same terms. For example, I found that the terms information organization and technical services were both used as broader terms for cataloguing. I ultimately decided to use the term information organization because I felt that it could better represent the other terms which I had grouped with cataloguing. Therefore, I ended up having cataloguing, indexing, and classification all as narrower terms for information organization.

I also looked through the thesaurus for ideas of what to do with some terms that I was unsure of, like evidence-based librarianship. So I looked up the term in the different thesauri. In LISA, the term evidence-based librarianship is placed in a hierarchical structure with Librarianship and Library and Information Science. I liked this arrangement and decided that Library and Information Science would be a good broader term for evidence-based librarianship as a whole-part relationship.

The last relationship type is the associative relationship. This relationship is formed between conceptually related terms. These are expressed in a thesaurus as related terms or RTs and are also reciprocal. The related term is one which can be used in addition to the basic term but is not however in that hierarchy. This can help the searcher find a wider range of terms which may suit their needs (Cleveland & Cleveland, 2001).

I found choosing related terms far more challenging than picking out broader and narrower terms. Most of the terms seemed to be related in some way or another but I could not make them all related for fear of overloading the thesaurus and making it too cluttered and confusing. I also found it a little challenging to discern between which relationships should be hierarchical and which should be associative. For example, I created the term research as a broader term for health research; however, I was torn over whether to also include statistics as a narrower term or a related term to research. Ultimately, I decided that statistics did not fit as a narrower term to research because it is not a type of research but is instead the end-product of research; therefore, I placed it as a related term.

While creating the relationships I ran into a few problems. One problem I had was that there were a few terms that really did not seem to fit anywhere and I was unable to think of any broader, narrower, or related terms to go with them. These were the terms which I was unsure if I should include them or not at the beginning of this project. I finally decided not to include them in the thesaurus. These terms included ideas, future, and 1990-2000. The term 1990-2000 seemed as though it would work better as an identifier than a descriptor and, therefore, should not be included in the thesaurus. The terms ideas and future seemed to me to be too abstract to include as an indexable concepts.

Another issue that came up while I was working on creating relationships was that I had to make some more decisions regarding compound terms. The terms seemed to be easier to handle once I had decided on a proper form. For example, I had to decide what to do with the term rural reference libraries, which I had decided to leave as one term in my original subject terms list. I could break the term up a couple of different ways or I could leave it as it was. According to Aitchison, compound terms which represent two different principles of division should be factored (2000). Following this rule, I decided that splitting the term up would not drastically alter the meaning and, therefore, chose to use the terms reference libraries and rural libraries. I did the same with the term reference and information services by factoring it into reference services and information services.

Previous   |   Next