acadamic research的比较 by ahyangyi

Google Scholar

• Indexed the most number of papers. The keywords "data race" yields ~1,960,000 results (but only the first 1,000 are viewable, evil!!). The paper ranked first in its result, "Eraser: A dynamic data race detector for multithreaded programs", has 114 versions and is cited by 829 papers. The papers isn't exactly sorted by citation numbers.
• Try its best to make it convenience for the user to actually get the papers. 7 out of the 10 papers in the first page are immediately downloadable. What's more surprising is that Google provides a "Help from Tsinghua" link whenever you use a Tsinghua IP and the paper is available from the Tsinghua Library. I speculate that Google has done similar things for lots of other academic organizations.
• Advanced search allows you to specify the author, publication time and the conference/journal of the papers. Additionally, Google allows you to search patents & legal opinions. (What does legal opinions do with scholar, anyway?) You can also choose from 7 different subject areas.
• To summerize, Google Scholar seems to want to be the best tool whenever you want to look for papers.

Microsoft Academic Search

• Surprisingly, MAS seems to be unrelated to Bing. Even the "search" icons look different.
• MAS organizes multiple types of information: papers, conferences, journals, researchers, academic organizations, research domains. Their relationships are rather close: whenever you look for one thing, say, a researcher, some small boxes occurs on the left of page which gives you some useful links from other types like co-author, conferences and journals.
• The keywords "data race" yields 2814 results. Unfortunately, the first and the third are both "Eraser: A dynamic data race detector for multithreaded programs", as one of them spelled "multi-threaded" while the other does not use the hyphen. The two duplicates sum up to 656 citations, slightly less than the Google Scholar version. Again, the papers isn't exactly sorted by citation numbers.
• Provides direct download link for 5 out of the 10 papers on the first pare for "data race". Two additional links goes to ACM library and you can find download link there (isn't this because I'm inside Tsinghua?). Another link goes to IEEE Xplore.
• Provides information for each researcher/conference/journal/organization. And provide graphs including co-author graphs and publication/citation diagram.
• To summerize, MAS seems to want to make a point by organizing various types of informations cleanly and tightly.

ArnetMiner

• For those who haven't heard of ArnetMiner, you may try it out at http://www.arnetminer.org/.
• Very distinct ideas about what an academic search engine can do: ArnetMiner seems to emphasize on people instead of the publications. Actually, I cannot find a "look for a paper" option at the homepage. Instead I can search for expertise, courses, social graph and conferences/journals. In addition, ArnetMiner produces ranklists and a "Topic brower" which tries to give you a extended list of all topics of academic interest.
• However, if you just type in "Data Race" at the expertise tab, you will actually find a paper list at the right-bottom corner of the page. But the familar paper  "Eraser: A dynamic data race detector for multithreaded programs" isn't ranked No.1 here. The No.1 paper is called "Data Mining: Concepts and Techniques", with a reference count of 10681. Perhaps the overwhelming reference count is the reason why ArnetMiner makes this mistake.
• Instead, you can find lots of researchers when you search "Data Race". Comes with two different ranking algorithms. And ArnetMiner provides lots of information for the researchers, notably social graphs and ability radar charts.
• To summerize, ArnetMiner focus on the social aspect of the academic realms.

BiomedExperts

• Focus on Life science. A bigger problem is that I cannot find a search box. Instead, it seems want to me to pay and login. So let's throw it away from our report.