I read a few blog posts lately about where the Web is going and they struck a cord. They managed to articulate more clearly some of the things that we felt intuitively when we started building ContextVoice / uberVU and were not able to clearly express.

It’s going to be a pretty long article, so you can skip to the short version/conclusion here.

Streams

The Web is definitely moving towards a stream-like, conversational structure. Twitter, Facebook, FriendFeed, all of these work as streams. More importantly, to emphasize the trend, feed readers and even Gmail are starting to look like streams/conversations. I find myself spending over 50% of my personal time online in a stream environment and, if you consider Gmail a stream-like experience, probably close to 80% of my work time inside streams.

There are two approaches to streams that, if you look at how we function as human beings in real life, are totally complementary.

Imagine meeting a friend outside a coffee shop en route to a meeting. The friend, with the context around her is an atom in a stream of inputs that demands your attention. You start talking to her about something, the conversation then suddenly focuses on something else. Your phone rings and so you focus your attention on the phone call, which is actually just another atom in the stream. Your meeting partner is going to be late. You then switch your attention to your friend and start talking about something completely different.

This is the way we operate in real life when we try to find information or socialize, yet the Web does not reflect that. Until now.

I’d associate the stream of information all around you with the way Twitter works, while the way we dig deeper into a certain subject / interact with a social object is more like FriendFeed works – or ContextVoice / uberVU for that matter.

Whatever the type of stream you experience, soon enough you realize you cannot consume it all, as it was not conceived for that. Navigation becomes more important that it has been so far with pages, because a stream changes continuously. It’s not just about finding some “relevant” information, like you do on Google, because that may not be relevant anymore in a changing stream. I’d argue that it’s not even just about finding “real time” information, like you do on Twitter search, because that only addresses the time axis. It’s a combination of the two that probably yields the best results.

How we think about streams

We started tracking comments from all over the Web around URLs because we felt a conversation does not stay trapped within a single site or domain and so you should not experience the conversation that way. The conversation flows between blogs, Twitter, FriendFeed and social networks but it’s essentially the same conversations.

Pages don’t matter anymore, it’s all about being able to tap into the right stream at the right time and filter it as you like.

What ContextVoice does is find related bits of streams (trackbacks, comments, tweets) from the larger stream, sticks them together and creates a new stream that we keep track of. We think it’s one way of navigating the larger Stream, by sticking together related pieces that are part of the same conversation.

As streams will be the fabric of the Web, being able to tap into streams to extract valuable information is going to be key. Navigating and participating in streams will be part of most web apps and will be used by both people and businesses. Everybody will be using this technology, just like e-mail and RSS. It’s already happening, with buzztracking tools and customer service on Twitter.

We can easily imagine applications tapping into streams to extract events, financial and health information, interests, relationships and a lot of other types of data, processing it and then pushing back to the Stream, to be consumed. This is where ContextVoice comes in providing applications with easily accessible ways of tapping the Stream.

Taking this to the next level, we believe search is probably going to still be the preferred way of finding information in the Stream. But I think of search not as it is today, but mostly as a combination of Tracking and Discovery. You’ll express what you want to know about, not necessarily now, but perpetually, and search system should be smart enough to pick up the appropriate pieces of the Stream, apply filters and deliver the right data to you, as a stream.

This is the approach we’re taking with uberVU. uberVU will be a search and analytics product, where you’ll be able to both search for conversations around URLs and search for keywords. The results returned will be whole conversations, not just posts or tweets containing the keyword. And, of course, you’ll be able to experience searching for a keyword as a stream of conversations updated in close to realtime.

We think this is an important distinction which ties into the role Context will play in the new Web.

The Role of Context

The Stream is made up of atomic pieces that constantly flow, with not much context to go by. Think of your Twitter stream – continuous updates on different subjects, shared links – each update does not have much context except the author and the time of posting.

People’s need for context became evident even on Twitter. That’s why we have @, RT or #. These symbols are trying to bring in a piece of context and encapsulate it in the atomic bit of information. So context is needed if we are to make sense of what information in the Stream means and where our attention should lie.

A more subtle implication of this fact is that context should be able to be incapsulated in the information itself and easily travel with the information wherever it goes. Of course, context changes over time and a piece of information can be looked at from within different contexts.

This is exactly why we built ContextVoice as a different product than uberVU.

Firstly, ContextVoice gets CONTEXT around stories from all over the Web. Instead of getting simple bits of information with no context around them, we get comments from all over the Web that are about the same story. This makes for some interesting information. You can see actual comments but also how they’re related to each other (RT, threads, comments to trackbacks), how they happened on the timeline, how fast the conversation has accelerated and decelerated, etc.

Secondly, as I said before, CONTEXT should be able to travel with the data. This can prove really tricky, as some conversations are made up of over 10,000 bits of information. In order to solve this problem, ContextVoice can return all the CONTEXT around a story in a single API call. We can’t encapsulate it in the data, as it’s not practical, but we have set it free and easily accessible.

Thirdly, we are not the only ones that will need this data. A lot of companies will use it, as we move more and more towards a more streamy Web. What we’re doing with uberVU is a single use case in a sea of possibilities.

Context can be used for many things. More context is usually better, but after a certain limit, I think context will only be useful in realtime search. People won’t be able to process it, but we’ll need it if we want to get relevant search results in realtime. As uberVU will mostly be a search product, we’re trying to get as much context around stories as we can, there no such thing as too much.

In Short

The Web is turning into a realtime Stream. We won’t be able to digest it all, so navigation will be increasingly important. One way to navigate the Stream will be through Search, but smart realtime search needs a lot of Context around the atomic bits of information if we are to find the truly relevant information we need from this everflowing, noisy Stream.

We built ContextVoice in order to get Context around stories from all over the Web. We get comments, mentions, reactions, tweets and other things, some of them in close to real time. ContextVoice is an API because of two reasons:

1. Context should be able to travel with the data. As we can’t encapsulate this much context within the data itself, providing it as an API call seems to be the easiest way.

2. As the Stream becomes the very fabric of the Web, being able to tap it will probably become like RSS or e-mail. Everyone will need this technology as part of their apps or businesses.

uberVU, the product that uses ContextVoice will try to tackle one of the problems of navigating this Stream – what you’d call Search and I’d call Track/Discovery. Finding fully contextual conversations (streams) to participate in, not just atomic bits that contain keywords.

Back to top

It’s been a long article, but if you’ve come this far, I encourage you to hang in there for a couple more seconds and share your thoughts.