Jim Hu wrote:
For example, the web service at Pubmed provide the abstract and links to full text (at yet another website) for a publication. My users would want to add things like: "This paper describes a resource that turned out to be useful for doing X" or "Figure 1 in this paper shows this thing that the authors didn't notice" or "The xxx gene described in this paper is also known as yyy; they were shown to be the same 10 years later" etc.
I have a similar problem. At http://runeberg.org/ I digitize old books, among them several encyclopedias. For the sake of familiarity, you can think about scanned books in Wikisource rather than my website.
In many cases an encyclopedia from 1889 is useful for knowing the population of Aberdeen in 1889. It could be nice to report what the current population is, but in some cases it is also important to point out that the reported number for 1889 was indeed wrong. But if scanning and OCRing one page takes 3 seconds and proofreading takes 3 minutes, how long does it take to check all the facts? Not knowing how this should best be addressed, it seemed like a stupid idea to digitize more old works that are full of errors.
When Wikipedia was started in 2001 and started to get off the ground, this became the obvious place to put information on the current and historic population of Aberdeen. The scanning of old texts no longer had to carry this role. It was really only in 2002 and 2003 that I got the energy to scan more works for my own site, and in 2005 I scanned this for Wikisource, http://en.wikisource.org/wiki/The_New_Student%27s_Reference_Work
Turns out Aberdeen's population in 1911 was 163,084, http://en.wikisource.org/wiki/The_New_Student%27s_Reference_Work/1-0016 http://en.wikisource.org/wiki/The_New_Student%27s_Reference_Work/Aberdeen but this bit of information is not linked to or included in http://en.wikipedia.org/wiki/Aberdeen#Population
So one problem still exists: From the scanned book page, there is no link to the Wikipedia article that provides more up-to-date information. The reader of the scanned page can of course use a search engine, and will often find the Wikipedia article. But is this really the ultimate solution? And even if the Wikipedia article is found, the other scanned pages that link to the same article are not found from there.
Should each scanned book page include a list of links to Wikipedia articles that are relevant for the page? Could such lists be compiled (or suggested) automatically?
Should Wikisource have a [[category:Aberdeen]] that collects all pages, chapters and books that pertain to this town? Today the English Wikisource has one [[Category:Works by subject]], but under this is a very small tree, compared to all articles in Wikipedia. There is no category for Aberdeen, but one for Scotland that has 15 links of which 4 are to articles in the 1911 Encyclopaedia Britannica. The 1911 EB article "Aberdeen (burgh)" is not among these four, http://en.wikisource.org/wiki/1911_Encyclop%C3%A6dia_Britannica/Aberdeen_%28...
Wikisource also has a [[Category:Ottoman Empire]] that contains four articles from the 1911 Encyclopaedia Britannica, one other chapter and two other works. But the corresponding category on the English Wikipedia has 56 pages and 12 immediate subcategories. Even the sub-subcategory Ottoman railways has 6 Wikipedia articles. On Wikisource there seem to be 6 mentions of the "Orient Express", but these are found through Google and not through links on the website, http://www.google.com/search?q=%22orient+express%22+site%3Aen.wikisource.org