What is Interest Graph? It helps discover people’s interest in things with data-mining approach. Interest graph is an online representation of individual’s interest with others including others interests. Simply it can be stated as the connection between people’s choices they make or the things which are common between people and the interests they share with each other. However, the interests of a person changes as time passes.
For Example: A few years ago I was a big fan of wrestling (WWE to be specific), but as time passed by I became more interested in following Snooker and Darts championship. Now my interest graph would show a strong connection towards Snooker and Darts rather than wrestling. Another way to see this is that my general interest is sports, as these interests are connected to the sports or lie within the sports category.
Interest graph used on the social networking sites help advertisers to analyze the market in general and identify the audience they should target for their product or services. Advertisers can use personalized messages for marketing with the help of interest graph. You must have noticed Facebook advertisements on the right side of your wall, these advertisements use interest graph to identify the target market.
Most of us know that Facebook has been able to bring in the social trust factor in advertisements. For example, you are provided with information about your social circle’s activity (what they are reading, playing, or even listening too). The newsfeeds algorithm of Facebook relies heavily on the recent activities rather than the relevancy of the newsfeed. For example: it might be very difficult to find a status of a friend in the newsfeed set just a few days ago. Even though you are very well connected to this particular friend, you still would face the difficulty to find his/her status.
Interest graph platforms are way more difficult than they sound. Companies which are developing the interest graph platforms need to bypass some fundamental problems or challenges:
Data Collection:
There can be two approaches to collect data. You can either state that you require the user to provide the information OR you can follow the user’s activity and find out the interests yourself (i.e. explicit and implicit).
Explicit approach may create hurdles for the users and it might popup the question in users mind that why should they provide detailed information about their interest while they have the option to search the product or service required directly. An example of explicit approach can be Pinterest.
Implicit approach might best be defined with the help of an example i.e. Amazon. It uses implicit approach of data collection. The site has it all, your purchase history, your search history, and it even performs different co-relations (i.e. the people who bought this also bought that, or the people who searched this item also looked for that item).
No matter how you collect the information, but remember that when a user has to input his interests, normally the platform tends to generate inaccurate interests profiles of the users. So, our suggestion is to use the implicit approach and collect the data according to the real interests of the users.
Noise Reduction:
On social networks, every action of user is considered as a signal. For example: if you “Like” an image of your friend, taken in a football stadium watching a game between Chicago Bears & NY Giants, this necessarily doesn’t mean that you like Football. There can be a number of outputs taken from this action of yours. The list may include:
- You like to watch Football.
- You like Chicago Bears.
- You like NY Giants.
- You simply liked the picture.
- You wanted to start a conversation. (and so on)
However, if you uploaded the image of yourself in the football stadium watching the game, it would certainly mean that you like Football, you either like one of the teams playing the game or maybe both of the teams.
The algorithms are very typical, so do not worry if you get it wrong in your first attempt. Remember, perfect systems are not generated in the first attempt they are modified or changed according to the requirements over the time.
Interest Graph:
Even after the noise filtration, it is hard to prepare the interest profiles of the users. The complexity of the system should reflect interests which are generated by all the user signals to create a reasonably consistent profiles of the users.
Platform API’s:
An interest graph platform means that you need to have API’s to allow the apps or games to access them, so that they can use these API’s to personalize the app or game according to their requirements. For this purpose you need to make sure that your API’s should be refined to such a state that others can understand it and use them for creating customizable app or games.
Closing Comments:
We will end the discussion with the statement that Companies which have successfully developed the interest graph for personalized or customized applications, to provide the relevant data and recommendations will be the future. However, they need to know that user preference varies with time, location and environment.
