All things 3.0

ExperienceOn Ventures is a Spanish company based in Barcelona, founded in June 2007 by two Spanish entrepreneurs, Carlos Gonzalez-Cadenas and Sergio Berná Niñerola, who share a passion for technology and its practical applications to make people’s life better.

Today we’re a startup composed by five persons, and we’re working on releasing the first version of our first service.

What’s the problem we aim to solve?

There is too much information, too many products and services out there in the Internet, and every day we have less and less time to check this huge volume of information or perform research for products and services. The growth of the Internet is explosive.

The current way of interacting with this information is by directly browsing through web sites and web pages, combined with the usage of search engines to find more web sites and web pages. The problem with that approach is that, for non-trivial things, you end up making a lot of search engine queries, and browsing through a lot of information contained in a lot of different web sites. The process is tedious and time-consuming. Many users simply give up.

What’s our solution for this problem?

We call them intelligent services.

Intelligent services guide us in the process of resolving a problem, simplifying the information review and decision processes. They should be capable of:

1. understand what you want, using an appropriate interface (text, voice, …), in your language, when you’re using full sentences or just some keywords
2. search, retrieve and analyze the information needed to solve your problem at machine speed, without requiring you to find and read it.
3. presenting you summaries of only the relevant information, filtering out the irrelevant and noisy information.
4. when there’s a decision to be taken, presenting you the available choices, with enough information for taking decisions.
5. making recommendations relevant to the concrete situation and your concrete preferences
6. personalizing the results, taking into account your unique goals and preferences
7. when requested, working in an unattended manner, that is, being able to perform background tasks in a continuous way, requesting our intervention only when it’s really needed.
8. helping us discover new relevant information or services that we may not discover in other case.

A simpler way of describing this: let machines do what they do well (searching and retrieving tons of information, products and services and analyzing them for us, performing repetitive tasks, doing massive data-crunching, …) and let humans do what they do well (evaluating what the best options are and taking decisions).

How are we going to do it?

The first question you can have is: ok, are you going to create a super-Google that solves the questions for all the problems?.

We think that this is not only very hard to do (even if, like Google, you have a ton of money, some of the smartest people in the world and a great technology), but also we think it is just not the best approach.

We believe that a more contextual approach will be better for several reasons. By a contextual approach, we mean that we create services that solve the vast majority of the problems in an area that is big enough to be meaningful for the end users (i.e. electronics, jobs, real estate, travel, …).

First reason, your algorithms are tuned to solve these problems, you can get better and more effective results and things are easier to mantain, improve and scale. We feel that if you have one-size-fits-all super algorithms, it’s easier to break some thing when you fix another.

Second, you can create user interfaces that are tailored to the problem you are resolving. The multi-page list of results is applicable for almost any thing you can imagine, but it’s clearly not convenient for the end-users in many cases.

Our first vertical service

The first vertical we’re targeting is travel. Our brand name for the travel vertical is befogg.com, inspired by the mythical traveller Phileas Fogg, main character of Jules Verne novel Around the World in Eighty Days.

Why travel?. We feel that doing travel research, organization and booking is a pain. The user experience is broken.

When you know exactly what you want, e.g. book a flight from Tokyo to Bangkok for New Year’s eve, it’s easy, you can go to Expedia or Sidestep (or your local equivalent), and you’re done in less than five minutes.

But you’re out of luck if you want to organize your skiing holidays in some place you have not decided yet, flying with a company you have not decided yet, staying in hotels you have not decided yet … you know. Doing all the research in Internet (from more generic things like getting ideas to more concrete things like deciding which hotels or what places to visit there) is a very complex, time consuming and tedious process, that requires performing many Google searches, visiting many social travel sites, reading many repetitive travelogues, trip logs or reviews, obtaining country and airline security information, weather information, … (of course, all from different service providers), looking for quotes in different electronic travel agencies. Most of the people simply give up and perform this research via traditional travel agents or by other means.

We want to simplify and streamline the process, integrating all these sources for you, and analyzing the massive publicly-available travel information that exists out there for you. We ask your help when we need your feedback, and finally we present you with a small lists of results and summaries of the important information, so you can be better informed in less time, and therefore take better decisions.

You will never need to review hundreds of reviews (unless you want to :)) when choosing an hotel. We do it for you and present you a summary, with our recommendations. You will never need to browse through a pile of prices (unless you want to :)). No more irrelevant ’special offers’ that are not special at all.

We believe that it’s not necessary to choose between simplicity and power. We believe that is possible to just choose both.

Why are you doing this now?

It’s the convergence of several trends that make this moment so appropriate for intelligent services to come.

The first trend is, as we said before, the massive overload of information, product and services we’re suffering. Not only is a real problem for users (having a ton of information coming from multiple sources), but it’s also a lost opportunity if we cannot benefit from this massive set of information that are out there. Using a travel example, just imagine the power of asking questions to an intelligent service that has access to all the travel knowledge and experiences that are published in the Web and can process these information at a lightning speed.

The second trend is that now is economically feasible to rollout services that have such requirements of computing, networking and storage. Also, the economics are going to improve greatly with the rise of multicore and manycore processors as industry standards, driving down the ratio of price per computing power unit at an accelerated pace. With more computing power at cheaper prices, we will be able to solve more complex problems, resulting in an overall benefit for the users.

The third trend is that, after two eras of the web, end users are getting more and more sophisticated. They demand and are more prepared to absorb new advanced services. New ways of doing things, new different services, new interfaces, … This trend will be specially pronunciated in people that have born in the Internet age, persons that won’t conceive life without the web.

What about the technology?

We will write one or more detailed posts about our technology later.

As an apperitive, we use a lot of natural language processing techniques (with a focus on statistical ones), machine learning techniques, information retrieval and information extraction techniques, and a lot of constraint-satisfaction techniques.

All of these things are programmed using two programming languages (one imperative and one declarative), and are running over an HPC infrastructure composed by a distributed computation platform that support MapReduce-style computations and nested data parallelism, a distributed parallel fault-tolerant filesystem, several datastore backends in order to store and retrieve efficiently the different kinds of data we’re managing (a relational database, a document repository (similar to Google’s BigTable and an associative database), in a cluster composed by commodity boxes running Linux.

More to come …

Stay tuned!. More about technology, intelligent services, Web 3.0, … in the following posts. Thanks for your attention!