Thursday, 9 November 2017

4 New things about Google Scholar - UI, recommendations, and citation networks


I'm actually a pretty big fan of Google Scholar, which in some ways is better than our library discovery service ,but even if you aren't a fan, given it's popularity it's important for librarians to keep up with the latest developments.

In any case, I'm happy to see that Google continues to enhance Google
Scholar with new features. These are some of the new features and things
I've learnt about Google Scholar lately.

1. Google Scholar's new UI 

The new interface is a lot cleaner, particularly when on mobile and most
of the changes aren't really major (e.g. replacing text of "save" and
"cite" with icons) but I miss the easy access to advanced search the old
interface had. 
Click the down arrow button to get access to advanced search in the old Google Scholar

In new Google Scholar interface, it now tucked under the "ham burger menu", where more people might notice it. 
On the plus side, very few people knew Google Scholar had a advanced
search or even how to access it , so overall it might be a ok trade-off,
though it takes two clicks instead of one to access the advanced
Also the change to Google Scholar doesn't seem to have affected link
resolvers, various extensions that rely on Google Scholar via scraping
such as Publish or Perish , Google Scholar button, so this is still a relatively minor layout change. 

2.  Get recommendations of related works of other scholars' works

Official change announcement. 

For a long time Google Scholar had a odd gap. As arguably the largest
scholarly index in the world, with perhaps the largest number of users
of any scholarly search engine, it was well posed to use all this data
to create a fantastic recommendation system. Add Google's famed machine
learning and it looked like a no-brainer.

But it was only in 2012, nearly 8 years after launch that Google Scholar added a recommendation system.
And as you might expect, the recommendations are excellent. While other
recommendation systems for scholarly material exist (e.g. BX
recommender, Mendeley's, various publisher based ones), none in my
opinion are as broad ranging or timely as Google Scholar for the reasons
already mentioned.

Google Scholar recommended articles

Still there was a curious gap. The recommendation system only gave recommendations based on the works already in your Google Scholar profile.

The flaw here is obvious, what if you were working in a new area you
haven't published formally yet? Arguably this is exactly when you have a
greater need of the help of a recommendation system.

 I wanted a feature where I could put a set of articles into Google
Scholar and it would give recommended articles over time. One crazy idea
I had back at the time was to create a brand new fake Google Scholar
profile , load it up with works of articles I'm interested in , keep the
profile private and leverage on the recommendations provided.

Unfortunately this doesn't work, because the Google Scholar profile has to be public before recommendations appear.

Another way that probably works is to exploit the fact that papers
deposited into ResearchGate, preprint servers do appear in Google
Scholar and hence can be added to your profile fairly quickly. So you
could example, create a quick working paper (with citations to works you
know of) and deposit it in a institutional repository or preprint
server that is indexed by Google Scholar. Add those into your Google
Scholar profile and wait for recommendations to appear. But this still
seems really forced and do you really want to mess up your profile just
to get a few recommendations?

So the new feature added by Google is much appreciated. While you still
cant add any arbitrary set of articles, you can go to any Scholarly
profile and choose to follow the author's new works, citiations and most
importantly articles related to the author's research.

Follow Harzing profile to get recommended articles similar to her research publications in Google Scholar

It's not super clear to me if it just
sends new articles via email or whether it updates the recommended
articles list you get within Google Scholar, I suspect the former and
technically articles shown this way are alerts not recommendations?, but
it's still useful. 

3. Google citation profile improvements - allows one time export to ORCID 

This isn't a new feature in Google Scholar but fairly new feature in ORCID.
I often find many researchers have their Google Scholar profile fully
filled up with their works (no doubt partly because Google makes it so
easy , particularly with auto or semi-auto updates and partly becuase
they reason the profile increases their visibility), but are reluctant
to spend the time to get their ORCID profile populated.  
Exporting works via BibTex
This of course only works as a one-time upload and you would have to
continue to update future works, hopefully by other automated ways (e.g.
via Journal crossref links, from CRIS/RIMs etc). 
Another fairly new feature in Google Scholar citations is that they now
try to group together authors by institutions. So for example when you
search for the name of an institution in Google Scholar, you get
something like this.
Searching by institutions in Google Scholar
Clicking on the link gets you this.
Top cited profiles from the institution
There's a study on how accurate this institution matching is  but what are the practical implications for normal librarians who aren't doing advanced bibliometrics?
For one it allows you to fairly easy get top 10, 20 etc cited authors of
your institution, to complement the lists you get from Web of
Science/incites or Scopus/Scival. 
You can't jump to the end of results to gauge how many authors your institution has on this platform. 

It's unfortunate that for this set of results, Google doesn't list the
number of results, and neither can you gauge it by looking at the number
of pages in the results list  and you can only go forward page by
page(see above). 
I don't know of a way around it , even if you alter the url parameters "&after_author "or "&astart=30" it doesn't work.

4. Scraping of Google Scholar to create network diagrams/ Bibliometric networks

It basically works as follows. First the system allows you to search via Google Scholar for papers to add.
Below I searched the term Open access, and then added some of the papers
into the system. You can of course search for specific papers by title.
Once you are done with a set of papers, you can click on "Check
Citations" and it will use Google Scholar's "search within citing
articles" feature to see if the articles in your set of papers are
It took me a while to understand how it worked but here's a specific example.
Say you have the following two papers
"Eysenbach, G. (2006). Citation advantage of open access articles. PLoS biology, 4(5), e157."
"Harnad, S., & Brody, T. (2004). Comparing the impact of open access
(OA) vs. non-OA articles in the same journals. D-lib Magazine, 10(6)."
The system will automatically go to say the Google Scholar list of citations for Harnad, S., & Brody, T. (2004) and using the "Search within citing articles" check to see if G. Eysenbach, is included. 
It will do this for all pairs of authors in your set of reference
articles automatically. All these searches are done in a popup window,
if the volume is too big , Google Scholar will throw up a captcha for
you to solve and it will then continue. 
You can then export a basic visualization of the author network which
shows coauthorships and citations. Here's my first toy example, using
papers I cited in a recent working paper.
It isn't too impressive probably because I don't have enough papers but
it does show the structure I expected with 2 main clusters - one around
LC smith (1981) (old paper on citation analysis for library collection
evaluation) and one around Eysenbach (2006) , a well cited early paper
on citation advantage. I would have thought they would not be connected
at all (particularly since I remove Eugene Garfield's seminal
publications) but they still seem to be linked indirectly.though an
author who cited both. 
You can export the network for further study into the open source Gephi network visualization tool and
I have spent some time doing so playing with more complicated networks
like publication & author networks, using modularity to identify
clusters of works. I'll probably blog about this in a separate blog post
next time, but for now I'm very intrigued.

How useful are such networks for researchers? 

Could doing such network graphs be useful for researchers, particularly
those new to the field to help them see how their research fits into
existing research, and see connections they wouldn't otherwise have
Could this be autogenerated from references of existing papers to help
the reader get a sense of where the current piece of article sits?
Can such network graphs provide improved recommendations (or does
recommendations from Google Scholar etc already implicitly do that?)
How big a network (or set of articles) is needed before this becomes
useful? e.g. Is this useful only for thesis with over 50 references (or
better yet include everything in your reference manager not just what
you cited)? Would most researchers find that these network graphs only
reproduce clusters they already intuitively know or provide some
unexpected insights?
In a future post I will talk about my experiments on these 3 scenarios
a) Visualizing networks between publications that cite my old 2009 article  
b) Visualizing networks between publications cited in my old 2009 article and newest paper
c) Visualizing networks between publications cited in an article not in
my field. (to see if it helps orientate me better in an area I'm not
familiar with). 
Would I learn anything from doing such visualizations?
Of course this idea isn't new, I'm guessing there should be research out there on this. 
Existing tools like web of science have limited citation map
capabilities and the newer incites and scival also provide mapping
capabilities though often at the higher level meant for research
On the free side, VOSviewer also provides the ability to visualization networks of citations. The newest version 1.6.6  actually adds the ability to generate networks not just from Scopus and Web of Science citations but also from Crossref.

VOSviewer 1.6.6 supports Web of Science, Scopus, PubMed and Crossref 
So one can generate similar networks using dois from VOSviewer.
Still I suspect scraping from Google Scholar might give richer results
due to the much large scale of Google Scholar compared to say Scopus.
Also given the popularity of Google Scholar has a discovery tool, one
might find relying on other tools such as Scopus to create networks
might risk missing too many works found via Google Scholar. 


Hope you found some of this useful.

It's good to see Google continue to improve Google Scholar, while we may
not know when Google might decide to abandon Google Scholar , the
recent spate of improvements are a good sign it won't  be anytime soon.

Posted by

4 New things about Google Scholar - UI, recommendations, and citation networks


  1. These ways are very simple and very much useful, as a beginner level these helped me a lot thanks fore sharing these kinds of useful and knowledgeable information.

    Eczema Treatment
    Psoriasis Oil
    Hyperpigmentation Treatment
    Herbal Tonic

  2. It’s amazing in support of me to have a site, which is useful in support of my know-how. thanks admin, you surely come with remarkable articles. Cheers for sharing your website page.Excellent blog here...
    Oil and gas ERP
    Human resources management software
    CCTV Camera Dealers in Chennai
    CCTV Camera Installation in Chennai
    RFID Solutions

  3. I like your way to describe every points and i get a better knowledge to your post. Thanks for giving plesant information.

    Hadoop Training in Chennai