Saturday, April 10, 2021

Google Scholar, Microsoft Academic, and Semantic Scholar Annoyances

Since I use all three services, I will try to compare their strengths and weaknesses etc. I will use following abbreviations GS (Google Scholar), MA (Microsoft Academic), and S2 (Semantic Scholar). More items will be added in the future.

  1. GS, MA, and S2: sometimes revise down the citations count for seminal papers by a few thousand or so within several months. E.g. noticed it today (4/12/2021) for "2014 Very Deep Convolutional Networks for Large-Scale Image Recognition by Karen Simonyan, Andrew Zisserman" it changed from a maximum of over 61,000 to now 57,000
  2. GS: does not provide a unique identifier for the articles (4/11/2021)
  3. MA and S2: The count of references per paper often deviate dramatically. GS does not publish a references count. This certainly contributes to messed up citations counts for papers!
    Example: Generative Adversarial Nets, a highly cited seminal paper: MA 33 reference count; S2 41 (2/25/2021)
  4. GS, MA, and S2: Their count of papers per author often deviate dramatically. Google Scholar does not even publish the number in a straight forward  manner. Example: Joshua Bengio: GS 1075 total publications; MA 973; S2 (S2 has three author pages for Yoshua Bengio) 838 (2/25/2021)
  5. GS: Lacks profiles for highly cited authors e.g. Emanuel Parzen (1919-2016). Google Search also does not link to Google Scholar profiles. MA and S2 do link to author profiles much more consistently (6/20/2021)
  6. S2: Unfortunately, S2 contains hundreds if not thousands of junk entries! I provide feedback on them as I find them. However, there are way too many junk entries. Junk entries have consequences: E.g. they often screw up citation counts of the paper itself and the references list of other papers (6/20/2021)
  7. GS, MA, and S2: All three have significant problems to distinguish book chapters from single research articles. GS search sometimes only returns a "[Citation]" as a research result (sometimes with foreign language characters such as Japanese, very odd)! Consequences: It leads sometimes to duplicate entries and probably also to messed up reference list for other papers (6/20/2021)
  8. GS: Does not provide a list of the paper references (6/24/2021)
  9. MA & S2: Only these two allow to sort the references of a paper by several criteria. However, you cannot sort by original order or by last name of first author (6/24/2021)
  10. GS: I noticed now several times that the Follow this Author function for new articles by author does not work reliably. It misses new papers. I speculate, but it may have something to do whether the author is the first or last author or somewhere in between. This is very annoying! (8/9/2021)

No comments: