New Home for Riya: Some Ideas; Pictures by Pictures How To

True to its community driven spirit Riya’s Munjal Shah is involving the community in revamping the strategic direction of the Algo startup – He is seeking Ideas for an All New Riya Home Page. Here are my thougts – Munjal is offering rewards for fully working solutions – Cool! Maybe someone more HTML/CSS/Webapp happy will find this and take this to closure. Good way to add flavour to the Idea/Execution debate He Ha!

At Riya, a typical ‘solution looking for a problem’ algo based tech startup, the strategic direction is evolving:

I Pre-launch phase( getting the algo work, working out the value proposition etc),

II Launch phase( community buildup, performance planning, hype and hoopla etc),

III Post-launch phase( over engineering/under engineering realizations, based on user profiling, realization that users needs to be conditioned more to fully use ‘recognition’ till then live with their old karma of ‘finding by search’). Thus search is the ‘All New’ direction. Hence this community exercise. Hence this post.

Cartoon Courtesy Hugh.

Here is my 2 cents on how to go about it. As mentioned by Munjal, Picture being worth a thousand words, a single text input box does not suffice. Especially if search is the USP. My idea is simple: Search Pictures by Pictures – just the way you search text by text. My pragmatics(highly weighed by Google’s success) goes like this: Google is a text serach engine, so a text input box works wonderfully well. Extending it to a picture search engine, picture must be the starting point for Riya search. In other words, Riya’s home page must facilitate users in entering picture(s) and their attributes. And then, let the Riya engine do the magic.

If you are with me till now, then the natural question is: How the hell do you make it easy to enter pictures? Will it work for the common non-geek? Even most geeks will have trouble entering a picture, even if there is ever an ‘easy’ method of entering one! Are you kidding?

My answer to the million dollar question in a word is:Templates. Templates or more intuitive sounding Riay Filters are simply a combination of:


1) Context/Folksonomy Tags - what made Flickr flicker,
2) Riya Tags - Training input/attribute that Riya Engine understands +
3) [Layers of Pictures] - Eg Foreground layer1: 2-3 persons, Background layer1: Mountains, Foregound2: Cola can, Background2: Kite flying.


The Layers of Picture above begs for explanation. It is a set of stock light weight line drawing icons that captures one aspect of the (Target) pictures. It must be possible to drag and drop as many into the picture input area. For example if you are interested in finding a picture with both Larry Page and Sargey Brin, you start off by drag and dropping 2 ‘celeb face’ icons into the picture input area.

AJAX can be used to fetch the working set of picture layers available for drag and drop – like the WordPress categories I am using right now. The most important things is to get a substantial set of picture layers trained by Riya and make it available as building blocks for Riya Filter Templates. The text analogy I guess is Keyword.

On second thoughts, all this drag and drop template building can be reserved for advanced( read geek) users in the Power Search page. The general public will be given a relatively simple hoop to jump: 1) Choose template(AJAX powered), 2) Enter text search attributes, 3) Optionally set some selection switches 4)Riya!

A good stock of well crafted templates, an easy way of selecting them, will make it easy for the Google conditioned to quickly get Riya conditioned. Of course, getting good results for the search is even more fundamental. You can count on the Riya community to grow the stock of well crafted templates/Riya Filters.

IMO, this whole template/Riya Filter rif raff goes well with what Riya is all about and in terms of its current feature mix. Riya is a unique mix of power of computing and power of people network. Riya Filter Templates will be the synergizer of the combination.

PS: Feel free to comment, mail me at labsji at Gmail for further clarification etc.
Update: I have added a small implementation detail here.