TasteMonster: Cape Town based software company breaks the mould on machine learning:
TasteMonster Photo: Julia Ledingham
There is a new recommendation system on the net and it’s fantastic. The unique intuitive system for streaming sites enables users to access tailored predictions and match shows and movies in real time for you and your friends.
Cape Town based developer Rick De Villiers has finally released his new predictions and recommendations system TasteMonster, for streaming sites such as Netflix, Amazon Prime, Showmax and iTunes. The site enables you to gain accurate, timely recommendations and predictions matched to your viewing preferences.
There is a growing trend worldwide for people to be choosing streaming media providers over traditional satellite and cable television. But, when faced with so much choice, how do you know what to watch without trawling through each streaming channel? Sure you can look at critic sites like Rotten Tomatoes, but these sites aggregate ratings and provide you with a general consensus. This would work if you considered yourself to be statistically average, but chances are that you are not.
Software developer Rick de Villiers explains: “If you have experienced Netflix or Itunes recommendations you’ll know how irrelevant they can be. Most have fallen back to simple genre-based or associative suggestions, for instance I really enjoy Star Wars but dislike Star Trek. The problem here is that my taste in Sci-Fi movies is quite selective and simple recommendations just don’t work for me and are not accurate enough.”
How do we create the perfect solution?
The natural human method for discovering recommendations is to take suggestions from people that we trust. We’ve all had that experience where we are socialising with friends and someone says “Oh wow have you seen that movie”. The problem here is trust. We don’t trust too many people with their recommendations.
Imagine if a system could expose me to 1000 people around the world who I trusted completely, imagine all the great discoveries that I’d make. That would be powerful. We know that collaborative recommenders are not new. The ACM RecSys conference has met annual for a decade with players like Spotify, Netflix, Amazon and Google all attending. But recommendation algorithms are all about compromise. Real-time systems are inaccurate and accurate systems require days of processing to produce results.
So if we were to make the ideal collaborative recommender, here’s what we’d want:
- Real-time: As I rate a movie I see my recommendations recalculate immediately.
- Predictions: I want to be able to find out what I would think of a movie based on all the feedback from the people I trust. I want a predicted score.
- What to watch with friends: The system should be able to match all my predictions to one of my friends and calculate our mutual enjoyment of any movie. This way, if we are trying to decide on something to watch together the system can help us choose what we would both enjoy the most.
- The system must be current: I want predictions on the latest content, not just old ratings.
In creating TasteMonster we have created a system that can do all these things.
- TasteMonster is able to process 25 million distinct ratings and generate 16 000 movie predictions in under a second.
- If you add your friends, every single page on the site will show you scores for both of you.
- TasteMonster also draws ratings from hundreds of critics. But unlike other sites, we don’t aggregate the critics, we actually profile them against you, so you are only getting recommendations from critics who share your taste.
This summary of TasteMonster is aimed at a general audience. If you are curious to know more about the technical details of TasteMonster we love to share that.
To experience the technology for yourself, go to www.tastemonster.com. If you have any questions or would like to discuss TasteMonster in more details, you can contact Rick on rick@tastemonster.com
CLICK HERE to submit your press release to MyPR.co.za.
Source: MyPR
Submit your own Press Release or get free content for your publication.
Video – How to post your Press Release on MyPR:
My Press Release (MyPR) provides free South African targeted content and press releases for on and off line publishers and media. Registering on MyPR enables public relations companies and practitioners to post articles under their own author name. Public Relations Companies and practitioners, event organisers, sports and political writers use MyPR as a distribution mechanism for their press releases, content, images and news. A forever free subscription allows publishers and media free access to all press releases, content and images on MyPR.
This press release is courtesy of WP Robot plugin – CLICK HERE to learn more about WP Robot for WordPress – a massive time and money saver.
