[ExI] Google Scholar

[Anders Sandberg]

On 2012-01-07 21:20, Natasha Vita-More wrote:
> Does anyone know how Google Scholar selects its material?

It most likely uses heuristic web crawling. It looks for material that 
either *looks* like academic text or is in an academic repository (which 
might also be deduced by some text mining algorithm). So it is not 
entirely reliable, but then again, academic journals are not entirely 
reliable either.

My own experience is that when casting your net widely about a topic 
scholar is useful. When you want to find high quality articles about 
something you already know roughly "where" it is dedicated search 
systems and indices like Pubmed and Sciencedirect are better. But when 
you want to see if there is *anything* about something, even in remote 
disciplines, then scholar is useful.

Overall, academic search is an interesting skill. In my job I often have 
to look up unfamiliar topics, often without even knowing what they are 
called. This leads to an interesting iterative search where I try to 
find topics closer to what I want, use their terminology to find 
something even closer, and so on. Once I get near my quarry I start 
looking at papers they cite (ever Introduction section in every paper is 
essentially a mini-review about the topic!) and papers that cite them. 
Following those links and repeating I try to find the core papers 
everybody cites or review everything, and it is those that I finally sit 
down and read properly.

What I really would like is a structure-based search engine that could 
find me theorems linking concept A and B, and using methods from domain C.

-- 
Anders Sandberg
Future of Humanity Institute
Oxford University