Will our online activities affect our real life social interactions? asks Corvida.

Well, we certainly think so. And this is precisely the topic of our next event Proactive Displays: Bridging the Gaps between Online Social Networks and Shared Physical Spaces, to be held at the Artificial Intelligence Research Institute in Barcelona, next Friday June 27th.

The Strands innovation team in Seattle, lead by our Principal Instigator Joe McCarthy, works on developing technologies that bridge the gaps between the digital and physical world.

The focus is on creating large displays applications that can sense the people and activities taking place nearby and show content relating to those people, activities and place. Proactive displays enable people to share the richness of their online lives with others in physical contexts in which they feel – or want to feel – a sense of community and connection, bringing the benefits of virtual communities into the real world.

Some of the possibilities envisioned by science fiction movies are a bit disturbing (see this provocative picture from the film Minority Report). However, we believe that if users have total control to opt in - and out - from being exposed to things that might be relevant to them, these technologies can be friendly and useful.

We’ll leave for another post a description of some of the applications of these technologies, its evolution and challenges.

Those intrigued by the growing trend of visualization techniques, will be interested in the responses of Justin Donaldson to a questionnaire received from the British publication Five Eight magazine. Justin is a MyStrands’ researcher, Indiana University PhD Candidate, author of the Artist Network Visualization and the Maps of Recommendations, and 1st place Award Winner in Netsci06:

1. Explain briefly the outline of your service and what it hopes to achieve.
The Artistnet Interactive Visualization (see video) generates a network model from our user’s play activity. This is a very similar process to what we do “behind the scenes” with our recommender engine, although on a much, much smaller scale. However, even in this limited version you can often see how individuals relate to each other through their shared music tastes. The network visualization makes this apparent when two people’s “strands” of artists become connected. Most likely, those two people have something in common with their musical tastes, and may find something new and interesting in each other’s recent music listening history.

2. What do you think music visualization services bring to the table over other recommendation services?
Music visualization services represent recommendation information in a much higher level of detail. They expose recommendation as a “field of possibilities” instead of a simple ordinal list of results. Other MyStrands visualizations (such as recommendation mapping) seek to do just this. Exposing the relationships between the artists, songs, or albums in this fashion can help an individual come to a better understanding of how the artists, songs, and albums relate to one another in a broader cultural sense.

3. How are they of benefit to consumers and music users?
They benefit consumers and music users in the sense that they offer a richer portrait of the music possibilities available at any given time. People that are passionate about music are then able to explore their options in a more involved way.

4. How important are visuals to users experience of music in an age where cover art has been cast aside in the digital realm?
Cover art is not “completely” dead. It is experiencing a rebirth of sorts in the form of small thumbnail images (such as the “coverflow” visualization in iTunes). There is no doubt in my mind that the resolution and multimedia aspects of “album art” will continue to increase well beyond its current primitive state, and will once again become a significant component to a musical offering. With high definition televisions and sound systems at reasonable prices, the time is ripe for a multimedia “convergence” of sorts which will most likely kick off with music. In a sense, this is already happening with music videos. In fact, the most popular videos right now on youtube are often music videos, and MyStrands has just started providing recommendation for these videos with MyStrands.TV.

5. Where do you see these kind of services going in the future?
I think we as consumers will have a heightened awareness of our digital entertainment options. Eventually, recommendation service users will split into two camps: Those who want “push” recommendation, similar to conventional advertising where options are provided to them in a straightforward manner; and those who want “pull” recommendations, where an individual can explore the world of media possibilities relevant to them, and eventually play an active role in spreading new media that they enjoy. This sort of behavior already exists through informal social interaction (”Hey, did you hear about this new band?”…etc). However, recommendation systems will seek to automate and optimize it for everyone: listeners, artists, labels, studios, and venues. It’s gonna rock.

By Dr. Rick Hangartner, Chief Scientist, MyStrands. Guest column published in MSearchGroove

No, the headline on this entry is not a careless grammatical error. Nor is the question really “What is the recommender market?” That implies that “recommenders” are mature, well-defined technologies that deliver specific features and value to the online world. Emerging recommendation technologies are currently setting the standards for discovery and personalization in today’s social networking-dominated web 2.0 environment, and the future of online social networking is all about discovery and personalization. While search engines help you find things you know you are looking for, discovery helps you find the rest.

If we accept that every business must make its case in 10 to 20 seconds on its Web site, then we are all but forced to admit that recommenders, more than anything else, represent the conceptual answer to the question: “How can I get that visitor/user/customer to realize that I offer something of value to him or her?”

Although venture capitalists and Web 2.0 users may find that claim to be just the tiresome excuse they need for hitting the “Back” button, the point is that a good argument can be made that unlike search engines, the idea of recommenders is a formal concept that has as many different concrete examples as there are separate market applications. The recommender industry really is the business of pulling three components together into a system that helps a user-driven business convince their potential customers that they should stay for a while. These three elements include:

1) An effective model that relates the needs visitors have to what the business offers.

2) Quality data to build a model instance that relates specific needs to specific offerings.

3) Unobtrusive means for easily and quickly determining an individual user’s needs.

Note that these three components are not quite as simple as “good (statistical) algorithms,” “a lot of data,” or “simple user interfaces.” In the coming years, defining an effective model will increasingly involve a scientific approach to understanding user needs and the market strategy of the business. Gathering quality data will require more sophisticated understanding of which data is actually relevant to the model. Devising means for characterizing an individual user’s needs will depend on a refined understanding of how people implicitly and explicitly signal needs that they themselves may not even fully understand.

In short, the recommender industry is the evolving business of building and deploying systems that reify some of the psychology of human economic transactions. What this means for the marketplace seems relatively clear: Search engines as we know them will never disappear. In the near term, search engines will increasingly incorporate simple recommender technologies to handle approximate queries (e.g. “You asked for this, and based on similar queries/behavior by others, you might be looking for this.”). But in the long term, the recommender industry will be larger, and recommender technologies will be more pervasive than the search industry and search technology as we know it.

Beyond that, some general themes about the future of the recommender industry that seem to be worth watching for include:

Multiple revenue models: Unlike search engines, which primarily are monetized through contextual ads of some form, recommender systems will be monetized in multiple ways. Recommender technology suppliers will continue to partner with customer businesses to derive revenue as a share of explicit sales increases directly accredited to the recommender system. In the longer term, recommender technology will increasingly enable business models, including advertising schemes, which could not exist without it. An implicit valuation for a specific application of a recommender system will be derived from the enabled economic activity.

Increasing focus on how users require change over time: In that recommender systems reify aspects of the psychology of economic transactions, there is an increasing appreciation for the probable value of responding to how economic behavior changes over time. This includes how an individual’s needs change over time and how the needs of the community evolve. The former can, in part, be accommodated by simply taking care to build a recommender system instance using data that is an adequate sampling of individuals whose needs are changing. Adapting to the latter may also require recommender system models that explicitly incorporate features of how community needs to evolve.

New concepts of personalization: One of the recent trends in personalization is using information about an individual’s social network to better characterize that individual’s needs and interests. This may be just one aspect of a new concept of personalization that puts the focus not on delivering an isolating, customized experience to a person, but rather on connecting an individual with affinity communities who can provide information of value to that individual. Few people really want to be out there all alone. And for those explorers who do, they might, in reality, be hoping to build a community of like-minded souls or be waiting for others to catch up with them.

More than anything, the future of the recommender industry is a business that will continue to grow and become more sophisticated as the science of recommenders greatly develops and to increasingly encompasses computer science, psychology, economics and cognitive science.

The old days when traditional gatekeepers programmed content for the mass-market are gone. Technology is shaking up the “chain of command” across popular culture. In the case of music, more and more consumers are looking at each other (friends, mavens, people they don’t even know…) to discover new songs and guide their tastes.

We have already talked about our approach for helping people discover things through other people’s tastes. Now we want to share further insight on how this process works.

Using OpenStrands (our public APIs), MyStrands’ researcher and PhD Candidate Justin Donaldson has developed a visualization technique (Artist Network Visualization) that shows in real-time how people correlate artists with other artists by listening to songs from their personal libraries. Strands of related artists will form from individual user’s listening histories. Eventually, these strands will connect, establishing the artist as a “hub” of shared musical information between the two users.

The visualization serves as a sort of primitive recommendation system as well. If you see an artist you recognize and enjoy, see who they’re connected to. You may just happen to see a new artist or track that you don’t recognize, and there’s a good chance that they’ll be worth a listen as well. MyStrands captures these opportunities in far more detail in our recommendation engine, enabling our community to explore artists, users, albums and many other items with equal ease. All these relationships form our “matrix of associations” which establishes a “ground truth” for how we at MyStrands understand the world of media content.

Try out Artist Network Visualization now!

This week MyStrands was very pleased to provide a video presentation to be used for a keynote address opening the SIRTEL 2007 (Social Information Retrieval for Technology-Enhanced Learning) workshop. This workshop was held In conjuction with the 2nd European Conference on Technology Enhanced Learning (EC-TEL’07) in Crete, Greece.

While MyStrands’ presenters Jim Shur and Rick Hangartner sat a stone’s throw from the foothills of Oregon’s coastal range, the workshop attendees were 10,448 kilometers away watching and discussing with them the video and its implications over Internet Chat. Thanks to Riina Vuorikari and all the SIRTEL organizers and participants for inviting us!

(We have uploaded the video to YouTube in four pieces (one, two, three, four), as we cannot upload the entire file at once. The entire video can be downloaded from here).

Abstract
Technology Enhanced Learning (TEL) represents an interesting new application domain for Social Recommender Technology (SRT). Many, and perhaps most, successful applications of SRT to date have involved recommending media items and consumer goods. For these domains, the knowledge base for the recommender embodies relatively superficial preference semantics, derived as relatively simple statistics from implicit or explicit preference data. It seems the recommendation tasks in many TEL applications will benefit from, if not require, knowledge bases that embody deeper semantics than just statistically evaluated preferences. This talk explores this issue of the role of semantics in SRT recommenders for TEL applications. We propose that the science of how people learn can guide us in building knowledges bases for SRTs applied to TEL, and for that reason TEL applications might be viewed as an evolutionary step for SRTs that places a new emphasis on the science of making useful recommendations.

Complex networks of human-generated links between items are the best mechanisms to analyze how our society considers and uses the products it creates and consumes. Although the resulting network is immense, it is a flexible tool for analysis and recommendation and is the foundation of MyStrands recommendations (read more).

This image is a way of visualizing a subset of UMA, our Ultimate Matrix of Associations.

We are excited to announce that Claudio Baccigalupo (PhD student at IIIA-CSIC) and his thesis supervisor Prof. Enric Plaza (IIIA-CSIC) have been awarded the “Best Case-based Reasoning Application Paper” by the 7th International Conference on Case-based Reasoning (ICCBR’07) for their paper “A Case-Based Song Scheduler for Group Customised Radio“.

MyStrands is sponsoring Baccigalupo and Plaza’s research, focused on recommender systems within a musical context.

Their current project is about a novel Web radio architecture called Poolcasting in which they are implementing and combining several different techniques (CBR, Pattern Mining, Group Satisfaction).

Congratulations to Claudio and Enric, and also to the IIIA-Artificial Intelligence Research Institute, this is the 3rd time IIIA has received Best Paper in CBR in the last 5 years.