Tuesday, December 4, 2012

The Google alternatives to ResearcherID and Journal Citation Reports

In a previous post, I explained how to use ResearcherID. In my post on assessing publication records, I explained what an h-index is and mentioned how many people judge journals by their impact factors. ResearcherID, which you can use to automatically calculate your h-index and other metrics, and the Journal Citation Reports (JCR) are run by Thompson Reuters, the first as a free service, the latter as a subscription service. However, Google has recently introduced its own variants of these services, and both are free. This post examines how they differ from the Thompson Reuters products and explores their advantages and disadvantages.

When searching for literature at work, I have always preferred Google Scholar (GS) to other publication databases, simply because it provides more search results, papers that aren't included in Web of Science or PubMed but also more versions of each individual paper, thus making it easier to find a site where I can access the article instead of merely admiring a paywall that my institution may not have paid for.

There is a trade-off involved here, however. GS gets more true positives, but also more false ones and more noise. See the screenshot below for a nice example. A formal study, admittedly from four years ago, found the same (Falagas et al. 2008): "Google Scholar, as for the Web in general, can help in the retrieval of even the most obscure information but its use is marred by inadequate, less often updated, citation information." This is perhaps one of the reasons that GS is still used less often for internal evaluations, but that may change in the future.
Garbled search result from Google Scholar: Candollea is the name of a botanical journal but the search engine interpreted it as a paper title. It displays this result although it does not even have any wrong, much less correct, information on the journal name, volume, pages, etc.

Of course, that does not mean that Web of Science is free of errors. About three years ago, I noticed that an unrealistically high number of citations was registered for one of my articles although several of the supposedly referring papers did not actually cite it; they weren't even from the same research field. (I informed their customer service of the problem and they cleared it up very quickly.) Even now, two of my articles appear in my profile twice, one time as the online PDF and one time as the print version. Interestingly, citations are only counted for one version in each case.

Note the new buttons "My Citations" and "Metrics".

Google Scholar "My Citations"

If you are logged in with your Google account (as you would use for blog-commenting, Gmail or other services), the counterpart to ResearcherID can be accessed with the the button "My Citations" on the Google Scholar website. To create a researcher profile, you only need to enter your name to allow GS to search for publications. You then mark those that are yours and add them to your profile. GS will subsequently show the number of your citations and your h-index - both likely a bit higher than in ResearcherID - whenever you access your profile. Beyond that, you can chose how much more you want to do. You can keep your profile private, only for your own benefit, or you can make it public. You can add a portrait picture, your research interests, your institutional affiliation, or a link to your website.

Finally, you can register your institutional e-mail address. This is relevant because there are two ways GS can update your publication list: automatically, which is recommended by Google, or by asking you for confirmation, which is what I would strongly recommend due to the aforementioned tendency of GS to retrieve some garbled or duplicated entries. At any rate, I am not sure how you could trust a machine to find out which publications are yours if your name is a common one. If you chose the confirmation option, GS will apparently send you an e-mail whenever it finds another of your publications that it wants to add to your profile, but as I only very recently created my own profile there I have not yet experienced this. Update: Works fine!

Google Scholar Metrics

The metrics button on the GS starting page leads to Google's counterpart to Thompson Reuter's JCR. Again, this has only just been introduced and is probably a work in progress; at least there are some curious limitations at the moment. You can select a field of science, and then narrow it down to a subfield (e.g. Life & Earth Sciences: Botany), but in each case the subsequently displayed list is limited to the top 20 journals. This is a bit strange as surely it would not be particularly difficult for Google to retrieve all of them and provide a "next 20" button somewhere in the right lower corner. As it is, if you want to see the stats for a more obscure journal, you have to enter its name in the search bar at the top of the site, but this makes it considerably more laborious to compare mid- to low-tier journals for the purposes of deciding where to submit a paper.

The second limitation is that as of writing this the only metrics that are displayed are the h5-index and the h5-median. Again, for an explanation of h-indices, see here; the "5" means merely that the calculation takes into account only the last five years. Now I understand that Google would not want to simply copy the JCR's problematic signature metric, the two year impact factor, but I am not convinced that the h5-index actually makes that much sense for journals either. The major problem is that its upper bound is the number of articles that have been published in the journal. Now for an individual researcher, that may make at least a certain amount of sense. If somebody publishes 25 articles in five years and somebody else publishes only 5 in the same time, it is probably okay to reward the greater productivity of the first researcher with the possibility of achieving a much higher h-index. But in the case of scientific periodicals, this plainly penalizes monograph series and journals publishing few long articles and conversely rewards journals publishing lots of short articles even if the former have the same or even more average impact. Because of this I would like to see more stats for the journals, such as mean of five year citations over all articles for example.

Well, perhaps the future will deliver. For the moment be aware that there are Google alternatives to ResearcherID and the JCR, and that you may want to check them out.

Reference

Falagas ME, Pitsouni EI, Malietzis GA, Pappas G, 2008. Comparison of PubMed, Scopus, Web of Science, and Google Scholar: strengths and weaknesses. The FASEB Journal 22: 338-342.

No comments:

Post a Comment