The art of the (exact) search

research data bites 08.

Key takeaways:

• For terms with multiple meanings or specific contexts, using exact search functions can help you avoid irrelevant results.

• Search is an iterative process

Last week we explored the new search query tool by Dimensions, let’s try a more traditional approach this week.

When exact isn’t quite exact

By default Dimensions’ exact search is not “exact” — for instance searching for “Hansen’s disease” is the same as “Hansen disease”: see ("Hansen’s disease" NOT "Hansen disease"), which returns zero results and therefore shows that there are no more publications retrieved using “Hansen’s disease” than when using “Hansen disease”.

This is because Dimensions employs a vocabulary that treats variations like plurals, possessives, and common spelling differences as equivalent terms. This simplifies greatly searches, as it means we do not need to account for all variations.

For instance, Mearns et al. studied type 2 diabetes, and used the following Boolean search (29 terms):

trial OR trials OR before-and-after study OR before-and-after studies OR cohort OR comparative study OR comparative studies OR Controlled OR evaluation study OR evaluation studies OR follow-up study OR follow-up studies OR interrupted time series OR longitudinal study OR longitudinal studies OR non-randomised OR nonrandomized OR nonrandomised OR nonrandomized OR non randomised OR non randomized OR program evaluation OR programme evaluation OR prospective study OR prospective studies OR quantitative study OR quantitative studies OR quasi experimental OR repeated measures

which in Dimensions is simplified as (19 terms):

trial OR “before-and-after study” OR cohort OR “comparative study” OR Controlled OR “evaluation study” OR “follow-up study” OR “interrupted time series” OR “longitudinal study” OR non-randomised OR nonrandomized OR nonrandomised OR nonrandomized OR “program evaluation” OR “prospective study” OR “quantitative study” OR “quasi experimental” OR “repeated measures”

Here is how I simplified it:

• removed standard plurals (trials, studies, ..)

• removed the s/z variations, apart from nonrandomized, as it’s not part of the vocabulary

• removed the programme/program variation

• removed “non randomised”, as it’s equal to “non-randomised”

Although I suspect we could add terms such as “quasiexperimental” (you would be surprised the misspellings found in scientific publications..)

PubMed automatically searches for plural forms and British and American variations, but not in the multi-word terms. So the authors could also have removed trials in the same way they didn’t use cohorts.

When true precision matters: full_text_exact

However, I recently encountered a challenge while searching for information on noma disease in Dimensions. When searching in the database, I realised it could be confused with Non-orthogonal Multiple Access (NOMA/NoMA), a 5G technology, but also Neonatal Oral Motor Assessment Scale (NOMAS).

To ensure I was retrieving accurate results, I had to use the following specific search query:

noma NOT full_text_exact:NOMA NOT full_text_exact:NOMAS

This query excludes articles related to NOMA technology and NOMAS, allowing me to focus solely on the medical condition.

In a less important topic, I plotted below the trends for color vs colour in full text, as well as color AND colour. Not sure exactly what happened in 1981, but likely a data artifact.

Guidelines for developing effective Boolean searches

In light of the previous discoveries, here is a very short guideline to search in bibliometric tools:

• Define key terms: Identify the main concepts of your research and list alternative terms, synonyms, and relevant variations (e.g., "type 2 diabetes" vs. "diabetes mellitus, type II").

• Do not introduce variations: Check for those that are less common by using “term1 NOT term2” to ensure there are zero results

• Use quotation marks for phrases: Enclose exact phrases in quotes (e.g., "before-and-after study") to ensure precise results.

• Apply OR for similar meanings: Combine synonyms or related terms with OR (e.g., "leishmaniasis" OR "kala-azar" OR "Leishmania") to capture all relevant variations.

• Use AND to narrow focus: Link distinct concepts with AND to narrow results, ensuring each concept is included (e.g., "type 2 diabetes" AND "longitudinal study").

• Leverage exact search for precision: For unique or ambiguous terms, use exact search functions (e.g., full_text_exact) to retrieve specific meanings or context (e.g., "mediums" for spiritualists, not media).

• Review and adjust: Test your search and refine terms based on results, adding or removing terms to balance specificity and comprehensiveness.

Happy searching!