Research finds Google Scholar better than Web of Science

The Utility of Google Scholar When Searching Geographical Literature: Comparison With Three Commercial Bibliographic Databases   
Journal of Academic Librarianship 

Volume 41, Issue 3, May 2015, Pages 322–329

Simona Stirbu, Paul Thirion, Serge Schmitz, Gentiane Haesbroeck, Ninfa Greco 
 
This study aims to highlight what benefits, if any, Google Scholar (GS) has for academic literature searches in the fieldof geography, compared to three commercial bibliographic databases: Web of Science (WoS), FRANCIS (multidisciplinary databases) and GeoRef (specialized in geosciences). 

Discussion and conclusion

With reference to the initial questions asked in this study, our results show that:

1)

The statistical approach applied to the data consolidates our hypothesis that GS leads the other tools widely on the number of results, independent of keyword, subfield, year of publication, or time of search. In contrast, FRANCIS yielded a very low number of results, leading us to deem it as the least useful of the tools in both human and physical geography. The interest of this database is possibly confined to more local and/or non-English literature. This database is multidisciplinary and covers broad subject areas in various languages at an international level. FRANCIS provides indexing, in addition to abstracts of books and articles from 2300 European-language journals, which is less than the number of journals covered by the other databases that we examined. A good percentage of the papers found in FRANCIS are multilingual, which might explain the poor performance of this database in our research.

2)

Based on the case study (of “urbanization” and “sedimentation”), it appears that GS remains on top even after removing duplicates and selecting geographical references (noise treatment). It is worth noting that GS's noise rates are similar to or even lower than those of WoS, the other international, multidisciplinary tool assessed. Both tools provided one valid reference out of two. In contrast, the specialized DB GeoRef always retrieves almost 100% geographical literature.

3)

With regard to overlap, we found that WoS is highly covered by GS, with values (66.5% and 82%) close to those found by Neuhaus et al. (2006) at 77%, and by Mikki (2009) at 85%. With respect to the specialized DB GeoRef, although GS's recall performance is reduced, our values (56.7% and 62%) are similar to the most recent ones (55%) found by Musser (2008). That strengthens the argument that GS's coverage has improved since previous work investigating this point (Neuhaus et al., 2006) with 26% coverage.

4)

By retrieving a large number of unique references, GS may yield additional information. Like GeoRef but unlike WoS, GS offers a broad range of references (journal articles of course, but also additional types of literature such as reports, theses, conference proceedings and books/book chapters). This additional literature can be very useful to researchers, and can obviously be acceptable as academic literature. Examples of titles present among the unique hits provided by GS are: IAHS PublicationsProceedings and Reports SeriesAGU Spring & Fall Meeting AbstractsGeophysical Research AbstractsGeological Society of AmericaAbstracts, and EGU General Assembly Abstracts. Grey literature (such as reports and working papers) are frequently produced and used in geography research. Specific research programs, contracts and grants results generally yield this type ofdocuments. These documents could provide pertinent data to researchers, and complement the information found in the traditional literature. However, some scholars are selectively interested only in peer-reviewed articles and consider that the traditional DBs are the unique way to quickly access to all peer-reviewed articles. But the notable presence of grey literature and books or books chapters is surprisingly not the main explanation for additional hits in GS: the number of unique “journal articles” is impressive. Since it is not yet possible to filter results in GS that only show peer-reviewed references, and after the examination of the lists of unique references, we are currently performing a study to verify the hypothesis that a significant proportion of these articles are published in a greater number of international peer-reviewed journals (e.g., the growing amount ofliterature in English from publishers in Asian countries). This study is ongoing.

5)

The performance of the bibliographic tools has to be considered by subfield (human or physical geography). The statistical approach suggests that the number of human-geographical references is higher in GS, whereas GeoRef is clearly geared towards physical geography. However, in absolute terms, GS should be useful to researchers in both fields because of the large number of additional references it provides. However, our results also show that GeoRef provides a significant number of unique items in physical geography, putting us in agreement with Mikki (2010), who says that GeoRef, which is highly specialized in geosciences, provides titles and abstracts translated from non-English publications, which could explain the significant number of additional items.

6)

During the nine-month period of the study, the number of references generated by traditional DBs for the years 2005–2009 proved stable, with an occasional slight increase. In comparison, GS data encountered more variation, and a detailed analysis of the references could possibly reveal even more important turnover within the lists. Nonetheless, there is such a big distance in raw numbers of findings between GS and the traditional DBs that our results and interpretations would not be affected by such an analysis.