A visual, discovery-based approach to Search
With the excitement on Google’s big reveal
tomorrow today on their vision of the future of Search, I thought it might be timely to discuss an idea that I have been shopping around since August 2008.
It’s called Bubble Search. And yes, it’s also got balls.
What is Bubble Search?
Bubble Search is an intuitive, discovery-based visual method for displaying, sorting and selecting search results.
A user is prompted to enter search terms into a field. Once entered, the related categories are displayed as Bubbles around the term. The actual links and full results are located in their organic position at the bottom of the page.
Terms and categories can then be ‘combined’, that is dragged-and-dropped together. There is no ‘one-tree’ concept to explore but rather subcategories appear as the parent bubble ‘pops’.
It allows search terms to be saved in a ‘basket’ for later re-use and to be mashable. This recombination of terms is intended to reduce typing, and plays on the fact that often we continue to search on with terms from previous searches.
By changing the Bubble View to Domains and Related it also enables users to find content providers closer to the types of information we are searching for.
What’s it trying to solve?
The ‘aha’ moment came for Bubble Search when I realised how painful it was to search on my cell phone. Using the keyboard on mobile is a pain. So we want to avoid using it where possible. When searching sometimes we get a list of results that aren’t quite relevant to what we want, so we have to search again.
All that typing when we know we are so close to our answer is so frustrating, worse still when we have already used terms that ideally we’d want to save or re-combine them later.
So what we try to do is this: You type your search query once. Bubbles appear. You click through the bubbles, popping up and down through the categories. If your device supports it, you multi-touch drag bubbles together to merge into a combined search query.
If too many bubbles appear on one page, it then splits them across, giving you a dialog to swipe to the next set and highlight selected bubbles (changing colour with selection).
Once you are satisfied with your selection, simply click a button in the corner and the relevant search links appear. To modify your search, just change the bubbles your selecting. No need to retype your query unless you get lost (click and hold a bubble to move back up a level though)
Alright, here’s an example you may relate to: You’ve metup with friends and you are all feeling hungry. Time for dinner! Someone suggests to find something in the area. So you key in “restaurants near South Yarra” into your mobile. Placemarks appear on your screen and a list of results.
What if you could then switch to see bubble categories, that you can pop into for more defined subcategories. Pop, pop, and you collate the terms and store them into a basket. All this popping is making you hungry. By this stage you are determined to have Korean BBQ in South Yarra.
However a new challenger appears and says they are intimidated by beef ie. vegeterian. Easy change of plan: Simply remove the carnivore Bubble and drag in a vegetable. Oh, it has to be in Inner City now because we are meeting other friends there too. Don’t panic, pop-up to Location and click on area that’s relevant.
All this time you haven’t typed except for that first moment. And if someone changes their mind again, you can reload previous baskets of terms. Or tell them to eat at home by themselves.
That’s what Bubble Search does best. Here’s some other reasons why:
Not everyone ‘gets’ Search
One of the classic problems that search engines need to solve first for users is the terms they plug in. The geekier database-minded people tend to load a phrase with only the appropriate terms eg. “Prevent cupcake icing cracking”. Ordinary folk tend to load up irrelevant or ignored terms eg. “How do I prevent cupcake icing from cracking”
If we could see the actual used relevant terms when we search as Bubbles around the phrase, such as those from Google Directories which categorises the web, this may be helpful to correct user behaviour for next time they want to search on this topic. No need to guess key search terms. See all relevant categories and drill into the ones you want. Importantly, this also allows for discovery of related but excluded from the typed terms.
Understanding Relationships to Prevent Bias
One of the inherent problems with search is bias. It is an almost inescapable consequence of search engines that a list of results will display prominently what it thinks the user is looking for and place it squarely in front of the user.
Ordinarily it is difficult to ascertain the relationships of the potential links and which relate more closely to our cause. So we use algorithms to understand the user’s search history pattern and to gaze through related search or predicted results based on similar user behaviour. This is called relevance.
The best searchers are researchers. If we look at what researchers do, we know they like to use trusted, cited sources. They prefer to go back to the original author, the one that is highly regarded by their peers. So likewise Google and others rank trust.
Following this further though, researchers also do their best to explain any potential bias. That is, Author X has an association with Company Y. Understandably the search engines have difficulty explaining this without giving away the secret sauce of their algorithms, therefore risking being gamed.
What if we could visualise this though without giving the game away? Let’s say we use a search term that’s quite controversial eg. “homeopathic remedies”. Say if we changed the Bubble View to show By Domain, we may then see a network diagram of the links associated from a search result’s domain to related terms.
We could use several techniques to highlight particular elements: by colouring to determine the age of content, or size of the bubble to represent amount of similar terms from that domain.
We may then hope to understand what as users the sources of information that are the most authoritative source. We may see that the results we are looking for, those that are supportive of our worldview, and then search on to related domains or articles. Or alternatively if we find there is an opposing view to something, what networks it talks to so we can persue that truth. This also facilitates discovery of websites, not just single answers to questions.
This functionality could be applied to more esoteric applications such as web forums, using Bubbles to search by poster, by related terms, by cross-linked domains.
In summary: Algorithmically display ‘opposing’ views, those that have little relation to another subgroup.. or view like-minded communities in order to discover more content
It isn’t just purely web or mobile search that Bubbles can assist. Already you can see there’s different ways such a visual system can be used from less-structured web searches to structured, service or product based search.
I’ve also been exploring using the visual search and the basket-store could be used as part of a primary OS from file exploring to spreadsheets.
That aside it’s a fun, intuitive way of collating and sorting. It is also particularly conducive for multi-touch, AR and visual displays that are becoming more prominent. Plus let’s face it, as a kid you loved blowing and popping bubbles!
When can I see it?
As George Brussard would say, “When it’s done”. See you in 12 years!… Ok seriously, I have created early iterations, mockups. I registered the domain, bubblesearch.net back on 24th October 2008 which was recently put to rest.
Initially an Android application, I have registered as an Android developer under Bubble since December 2008. I recently decided to abandon it and work on a HTML5 application for cross-platform compatibility.
And yes, this is all well before Google Wonder Wheel (19 May 2009) and before Google displayed it’s balls (9th September 2010). Let me be clear though: I love Google’s products and platforms, so for me it is humbling to think that I’d be on the same wavelength as people I greatly admire.
The intention however is to be search-engine independent, as some of the terms and conditions are a bit restrictive and I want users to select the search engine they feel works best. By doing so, it also allows the freedom to develop specifically for service (eg. WhitePages or YellowPages, where this idea works best).
For now though, the most recent models are based on Google Places API.
Whew! Long post!
Thanks. It’s over two years in the making really. Do you have feedback? Or interested in seeing more? Leave some comments and let me know!
Oh and this is just one related idea to Bubble. There’s more: Stack & Sort. To come! :)