Google Correlate Draw, Finds Data That Fits Drawn Pattern

Used to be that you first had data. Then you did analysis to figure out the patterns and trends in it. Now you imagine the pattern and Google Correlate Draw checks if there is a search term that correlates to your pattern. This is awesome.

In case you can’t yet make out what the fuss is. Imagine how a person picks a dress to buy, they find a dress that fits them. Now reverse it, pick a dress and imagine finding a person who would fit it. See that!

Google Correlate Draw Examples

Note: the blue lines are what I drew and Google Correlate plots out matching data in red.

Downward trend

And guess which search term pops up? Windows Server 2003 Enterprise! And a near perfect correlation too 🙂 LittleSnapper

Upward trend

Ends up lots of people had parking tickets on their minds. Am sure city councils would have seen their parking ticket revenue increase in this period. Google Chrome

Sine wave-ish cycle

Curiously the term ‘adware free’ seems to have had a fluctuating fortune. Google Chrome

Further thoughts

Correlate Draw’s approach brings out interesting possibilities. The basic idea of search using something other than text has been attempted before. For example i)Google Goggles, search for images or ii) Voice interface to search term entry or iii) Shazam, search for songs based on recorded snippet, all try to think of search as an activity that transcends looking for “textual” information.

But what is intriguing about Correlate Draw is its ability to search, or should we say match?, for patterns in quantitative data. And that too using an intuitive drawing interface. That twist of expressing a search query as a drawing is what makes this so interesting.

Imagine this. What if an research analyst in a financial firm ‘draws’ stock price movement patterns and have the system bring up companies whose stock price correlates with it? What if every time-series data could be searched in this manner?

There is also the UX aspect. The variation that can be expressed in a drawing can never be matched in a regular search interface. You could have textboxes to capture certain terms, sliders to express value within a range, a drop down to capture a single item out of a set and so on. But the expressiveness in a drawing would be hard to beat.

Anyway, those are my thoughts. What do you think about Correlate Draw? How far can you push this idea?

Time to extend the idea of bookmarks?

What is Bookmarking – , the simple act of storing a URL, whether in a personal store or on a cloud, revolutionizes our online participation. From being passive consumers of information, the act of bookmarking turns us into collaborators in extending this information. Those that we liked, that evoked a reaction in us, something that we could identify ourselves with, something that we would like to keep track of are all transformed by the expression of our interest in it. Especially when the bookmarks are centrally stored on a server somewhere.

Current Services – Online bookmarking services are dime a dozen. I use del.icio.us for all my(user name maheshcr) bookmarking needs. StumbleUpon is another popular service.

What is wrong now – However most of these services allow bookmarking of URLs and nothing more. Of course the URLs themselves could point to literally anything that could served on the Internet. These include text, images, audio, video etc.  But the URL that points to rich media, say audio or video files cannot pin point a location beyond the start. Let me elaborate.

What would be ideal– How about a particular riff within a song that I want to highlight? Or a quote within a video? Currently the best that we can do is to mention the time when the point of interest starts, and optionally ends. So we say, ‘watch at 1:10mins’ or ‘hear at 2:29’. As much as these are useful the fact is that user attention is not a decently engineered mechanism, accuracy and consistency is seldom guaranteed. If we could delegate this task of highlighting something in a larger body of work, to a machine, we should be better off.

Some compromises that exist today – To be clear some of this is done in latest video players where specific time points are highlighted on the timeline. This though is done by the producer of the content and does not allow the user to participate by marking areas as they see it fit.

Google’s ‘answer’ to this problem – Google seems to have attacked this problem with its Google Elections Search Video Gadget. Where a certain video is mined for text by recognizing audio and the resulting content attached as metadata. This comes very close to the problem I have discussed but feel it does not go the full distance. More could be done.

Wrap up – Has this problem been solved already? If not, would solving this add value to how we consume, discover and remember online information? What do you think?

[Update: The very next day I posted this entry a new service gets profiled on LifeHacker called Splicd. It lets one mark a specific portion of the video within YouTube. Interesting service. Have not played in detail yet, will keep you posted on how it turns out.] 

Ambient Findability by Peter Morville

What a book! I have barely covered the preface and the first chapter and am already smitten.

There are two kinds of technical books

  1. Ones that give precise directions to accomplish a certain task, the scope is narrowly defined and immediate outcome is emphasized
  2. Ones that describe the thought process more than practical steps. These are invariably abstract, cover a wider tract of ideas and have implications that are far reaching

This book belongs to the first variety. The tone is set in the preface when the author asks, “How did you get here…how did you find this book?”, and goes on to probe the process by which one could pick this single book amongst millions others. And how would one classify this book, the author asks that you don’t do any such thing but instead exhorts you to read it first, for “aboutness lies in the eyes of the beholder”. And I say what an insight!

I think its a characteristic of all great work that it tends towards philosophy. Not an algorithm to move from point A to point B but ruminates on the journey from A to B and leaves nothing as being insignificant in the course of its intellectual sweep.

And I leave you with a quote at the start of the first chapter..

At the seashore, between the land of atoms
and the sea of bits, we are now facing the
challenge of reconciling our dual citizenship
in the physical and digital worlds
.

Hiroshi Ishii

MIT Media Lab

Content Relevance

Serendipity! I had been mulling about the notion of contextual relevance just yesterday in search results and I stumble upon this blog entry from James. Please make it a point to read that blog entry, what was a problem in my mind is clearly and way more technically better articulated than I could manage!