They sent me a survey and asked me some simple questions, but I don’t think they asked me the right ones, so I’m going to give a free-form review here. I think it’s a great idea, and presents some features not available anywhere else, but it’s missing some important content, and like everything Thompson does, it suffers from some useability issues.
The search engine is called Thompson Scientific WebPlus, and you can only access it through their Web of Knowledge service, which they should really link to from the page telling you so. It’s also closed in the sense of being a somewhat curated list, with Thompson-selected authoritative sources ranked highest, but to their credit they do have a “suggest a site” button.
If you can get to it, it’s actually a pretty standard Google-like interface, but with Tabs for Topic, Author/Person, Organism, Drug, and Gene, instead of the normal Web, Images, Maps, News tabs at google.com. The search results returned are then presented as normal search results, with separate tabs for only News articles, Blog posts, or Repository links. I really like the idea of institutional repositories having their own tab, so you can search for a gene and get only links pointing to pages in a repository. I also like the idea of blogs having their own tab, so you can see what people are talking about, and having the News on a separate tab helps you scan for press releases and see how (badly) science is covered in major media. Once you’ve executed the search, you can narrow it by TLD, to pick out only results from .com, .org, .net, etc.
I tried a sample search that I’m familiar with in each one, just to see how this would work in practice.
A topic search for “multiple myeloma” has multiplemyeloma.org and myeloma.org at the top of the list, as with a standard Google search for the same. The notable difference is that wikipedia is dead last in the WebPlus results. News and Blog tabs are present, but no repository tab is available.
A Person/Author search for “William Gunn” looks much like it does on regular Google, again with Wikipedia being conspicuously absent. I’m pleased to see that my blog is the first in the results, as it should be for my name, and other results from the same domain are grouped. There’s even a link, which is dead, to my Nature Precedings account on the repository tab. Where things get strange are on the blog tab. They only list my other, infrequently updated, nonscience blogs, including a link to a development site on which I had a robots.txt containing
User-agent: * Disallow: /, but no link to Synthesis. Certainly not expected behavior. They’re just as bad as Google Scholar always was with author searches, too, because they aren’t able to figure out that results for William and Wallace Gunn should be shown in a search for W Gunn. WG Gunn is even worse. Pubmed remains best at handling this.
An organism search for “mus musculus” gives results similar to standard Google’s results, again trashing wikipedia, but it’s the repository tab fails this time, not including links to Jackson’s MPD or MGI, though they are in the regular web results.
A drug search for Velcade works like the organism search, presenting some good links up front and strangely lacking a repository tab. The blog tab works more or less as expected.
A gene search for DKK1 works similarly as well. If there’s anywhere I would have expected a repository tab, it would be here, but no such luck. However, this is where the blog tab really comes in handy, allowing you to see who’s talking about your favorite gene. Given the incompleteness of the results in the blog tab, I’m concerned about coverage here, but they do have scientificblogging.org, scienceblogs.com, blogs.nature.com, and even friendfeed, as well as some really weird french splog results.
So let’s summarize:
Topic – Works as expected
Person/Author – OK for finding people, even if the tabs don’t exactly work, but for finding authors it gets a FAIL.
Organism/Drug/Gene – Works as expected, but tabs fail.
Overall, it’s a strong first effort, and I expect the results to improve as they tweak things. I do think it’s a mistake to all but exclude wikipedia from results. For all it’s flaws, wikipedia still does a decent job on most things, and usually provides authoritative links for further reading.
As I was finishing the survey about the service, they asked some questions about ResearcherID.com and about something that seemed to be referring to a search portal. I wouldn’t be surprised in the slightest if someone at Thompson thinks a 90’s Yahoo style search portal would be a good idea, but search isn’t about being a portal anymore. It’s about helping people find the best information from the millions of results for any given term. I would not like to see an igoogle or yahoo search portal, but rather a open search that widely syndicates results so existing filtering systems such as social networks can help promote the more relevant results.
Other improvements I’d like to see:
and, of course …
Rafael Sidi from Elsevier emailed me and would like to get some feedback on what we don’t like about the way Sciencedirect does things, so please leave a comment here or in friendfeed.
He mentioned apps built using a WoK or Sciencedirect API a la salesforce. I think that would be awesome, and particularly relevant to the discussion about reading, bookmarking, sharing, discussing, and writing applications on on Nature Network right now.
Hmmm…why do we need a closed, curated science search engine? Haven’t we already got Scirus, which is open and curated, and Google Scholar, and the MS version. A fully open, curated but not parochial SE for science would be a different matter, one that can handle chemical structures as well as tEx formula etc.
I think there’s plenty of room for competition here, and WebPlus has a neat idea with breaking out repositories and blog posts onto their own tab, but as you can see, the reality leaves something to be desired.
Another major innovation would be a usable “initial stemming” for author search, so that publications published variously as “William Gunn”, “W. Grady Gunn”, and “WG Gunn” all show up in the same search. It’s not my fault it’s that way, either. Journals have different conventions for this.
Agreed that there is plenty of room for competition.
The improvement that you mentioned are valid for other scientific information products/publishers too.
In terms of author names, this is a painful task for everyone until we all figure out a (author name) standard in the submission process.
Aha! I’ve been discovered! Wanna see how deep the rabbit hole goes?
anything specific you’d like me to see? I’ve checked all of this out already…
There’s also Mr. Gunn’s Friendfeed page, which is where I send stuff that isn’t worth a whole blog post. That service also pulls in links, bookmarks, favorites, etc from other sites I use, such as del.icio.us, a site which stores links you send it. I find del.icio.us better than bookmarking the link in the browser, and it’s also nice to be able see the same list from anywhere, so I don’t run into the problem of finding a site at work or away which I’ve only bookmarked on my home computer. Additionally, you can see who else has bookmarked the same link, which can help you find related information.
FriendFeed aggregates the stuff from sites such as del.icio.us, so that people who keep an eye on my friendfeed page can know what I’m up to online. Friendfeed allows me to ask questions or post links and get answers or comments from people who’ve chosen to watch my friendfeed page, who are, in theory, people who share similar knowledge or expertise. There’s also people who watch my page who aren’t from a science background, but who want to know what us scientists are talking about. Many of these people work for publishers, but some are journalists or marketers or other people just generally interested in science.
Friendfeed isn’t just science, of course. There are groups of people doing the same thing we’re doing, but for food, economics, politics, photography, etc etc. There’s even a make-up group.