Beyond or beside Google: Week 3 - Duck Duck Go
Another week, another search engine. In the fourth week of April, I have tried out Duck Duck Go as primary search engine. In the line-up of search engines that I'm trying out, Duck Duck Go and Cuil are the newest and also the most different from the mainstream search engines. Although I admit to dreading Cuil week, in Duck Duck Go's case these seem to be positive differences. DDG takes some getting used to and it lacks some of the extra features offered by the major players, but once you do get used to it, it works well and is actually fun to use. So much fun, in fact, that I ended up making lots of notes and this may be one of the longest posts I'll ever write.
Duck Duck Go is the brainchild of Gabriel Weinberg, who operates the site by himself and he appears to be quite willing to listen to user feedback. When was the last time you posted a comment with some suggestions about a search engine on a blog, got a reply within hours and found one of those features implemented a few days later? Or received a personal e-mail response in minutes, rather than days or weeks after you submitted the feedback form? In fact, as I discussed my intention of doing these posts with him beforehand, I'm reasonably certain Mr. Weinberg will be one of the first readers of this post, something that I don't imagine any people for the other search engines will be doing. That, and things like the sense of humour demonstrated by the Cuiler and Cuilest spoofs are things I find charming (tip: search for "Google" on either of those). Still, the major issue with a search engine is whether it gives good and relevant results. Can you trust a small, independent search engine to give you those? As it turns out, you can!
The very first thing you'll notice when you visit Duck Duck Go is its slighly whimsical appearance. From the logo shown above (which actually changed recently, so I had to redo my graphic) to the buttons underneath the search field: Search, Information, Shopping and "I'm Feeling Ducky". Some people may find that and the somewhat silly name a little off-putting and it is of course highly subjective. It certainly causes people looking at your screen to take notice. Another way in which DDG is very different from other search engines is that you don't get a number showing you how many pages have been found for your query and doesn't split them up over multiple pages, instead loading further results into the existing page as you scroll down. Each result even has the favicon for the site it comes from displayed next to it, which is a nice touch and particularly useful to quickly identify authoritative sources without needing to read the snippets. It took me a while to get used to these things, but I ended up liking them very much. There are some niggles in the interface that I'll mention later, but on the whole it's uncluttered (ad free, too!) and easy to use. It is also one of the few websites I've come across where the interface caters for users like me who actually prefer to use the keyboard over the mouse. It is also quite likely the most most customizable search engine you'll find, allowing you to actually customise the search interface, rather than just add a Sudoku widget to your iGoogle homepage.
The very next things you're likely to notice is the zero-click info and the disambiguation pages. If the search query is one of many recognised topics, there will be a distinct box at the top of the results page with a quick description, often an excerpt from Wikipedia, which has been indexed at the paragraph level. Last week, I complained about Google not distinguishing between the different meanings for certain terms. If you type milk into Duck Duck Go, it'll ask you whether you mean the drink, the politician or the movie of the same name... or one of several songs and bands etc. In fact, I guarantee you that for almost any term, you'll find more meanings than you thought existed. Click on any one of those and you will find that the results have even greater relevancy.
Speaking of relevancy, I had no major complaints about the relevancy of the results returned. There were very few cases in which I didn't find what I was looking for among the first handful of results or where I needed to tweak my query to get the responses that I wanted. The number of results returned tended to be much lower than with any of the other search engines. You may find that DDG gives you only a few dozen results, when Google tells you that it's found 385,000, but there's this old saying about quality over quantity and that is true. Although it is said that the importance of the "long tail" should not be underestimated, I find it hard to believe that anybody looks at more than a few dozen results, perhaps a few hundred in extreme cases. If what you're looking for isn't there in the first few pages, then it either doesn't exist or you need to refine your query. There's no way the results on page 1,698 will be more relevant than those on the first handful of pages. Even if the amount of results is too limited for your tastes, Duck Duck Go will helpfully present you with a link to your search on Google when it runs out of results from its own index.
In fact, DDG does not shy away from using other sites for search as well, though the !bang commands. Want to search on Wikipedia? Just prefix your query with !w. There are many, many different different such keywords and they will carry out your search on the appropriate site directly, rather than having DDG's own result pages. It may seem like a minor thing, as you could just as easily select the appropriate site from your browser's search box if you have added it there, but I find it very convenient. Other such conveniences are things such as doing a search for a HTML colour code, showing a box in that colour at the top of the page and regexp matches (against a string, not against the search results). These last two were actually added in the week after I tried using Duck Duck Go, but I found them interesting enough to try them out when they were announced.
While most of the search results are highly relevant, I occasionally found myself missing a few. A good example was when I was searching for a specific article written by Raymond Chen. Perhaps I should have searched for notepad.exe site:blogs.msdn.com and this does indeed give the desired article among the first three responses. For some reason, as I often do when searching for one of Mr. Chen's blog articles, I prefer to include the title of his blog, thus using the terms oldnewthing notepad.exe and old new thing notepad.exe. While these gave me some related articles, the particular one I was looking for was not included. When actually including the page title, I would have expected a few more responses from that very site. Indeed, Google does that, giving me the specific article in the top 2 for both terms and has a handy link to "show results more from blogs.msdn.com".
Being smaller than the other search engines in my selection also means that in some cases, content may not be indexed at all or simply take longer to be indexed, especially that from smaller sites or non-English content. I won't hold it against Duck Duck Go that my own site doesn't rank highly in the results, but it would have been useful to get more timely content in my native Dutch in the rare cases that I actually did Dutch searches on current events. Trying the results now, a week later, already gives better results. Another feature that I found somewhat missing was an image search. This is largely mitigated by having the !image bang, which defers to Google for image search. That is in fact probably preferable to having a less than perfect homegrown solution.
Although I have had no cause at all to complain about the relevancy of the search results, there were a few things that struck me as odd. For instance, searching for the website of a restaurant in the city where I work, I used the query Het Kabinet, Breda, which resulted in an English PDF version of the menu as the first result. That was of course, a very useful result, although I would have expected the restaurant's homepage to show up first. Since DDG seems to prefer not having multiple results from the same domain, that page in fact did not show up at all, even if it was easily inferred from the URL for the menu. Oddly enough, entering the query as Het Kabinet Breda, sans comma, did make the home page the first result. Although I'm all for including punctuation in search, as I mentioned last week about searching for operator, this was a result that I did not expect.
Extracting snippets from pages, as well as titles from non-HTML documents is a very inexact science. This becomes very obvious when doing a site: search without any specific terms. This allows you to see what documents have been indexed on a site, but there is no term that can be highlighted specifically. Duck Duck Go gives some unusual page titles (which does not happen in regular searches that yield these documents in their results). For example, the indexed result for my list of DSP commands shows up with a very strange titled:
<!DOCTYPE xhtml PUBLIC "-//W3C//DTD XHTML 1.0 Transitional ...
length means the number of bytes, as opposed to samples. 10h - Output...1), hi (length-1) 1Fh - 8/2 ADPCM auto init DMA playback with reference byte...
shdon.com/files/dsp.txt - Similar Sites
Such (admittedly unusual) queries will also show HTML tags in a snippet that aren't there in the original source, especially in URLs. This snippet taken from the results for site:mozilla.org for instance:
Mdn - Mdc
developer.mozilla.org - Similar Sites
One final nit I can pick is that sometimes the special results aren't very obvious. The recently added colour search shows a bar in the selected colour, but no text label for it to point you to the fact that this is indeed the specified colour. At first, I overlooked it as being just a horizontal line in the interface. This also happened to me with certain calculation results. The Google calculator icon and large boldface font point out the calculation results clearly, but on DDG they can either be shown in rather small and easily overlooked type (search for 15 + 17) or as a delay-loaded zero-click info box (search for cos(0.3), which points to Wolfram|Alpha), which can be a little jarring. There seems to be a slight bug in the parsing of the Wolfram|Alpha results too, as a search for sin(30) gives an empty result.
With the exception of that last one, those niggles are fairly contrived and will have little if any impact on common searches. I had to search pretty hard to come up with them. This was a pretty long article, as there were just so many cool things that I came across. I haven't even mentioned the fact that DDG is privacy conscious, offering an https version and not storing IP addresses at all, which is what sparked my search engine switch in the first place. In all, I thoroughly enjoyed using Duck Duck Go for a week and I wouldn't be surprised to see it do extremely well in the showdown I have planned for the last week of my experiment. I am definitely considering making DDG my search engine of choice more permanently when the experiment is over.
- Week 1: Teoma
- Week 2: Google
- Week 4: Yahoo!
- Week 5: Cuil
Thanks for the info. I recently began using duck duck go and it is a great alternative search engine.
Post a comment