Introduction

TagMyBits is a flexible way of structuring the data held in a content repository by making the relationships between individual entries the basis for organisation. These relationships can be used to simplify the creation of recommender systems and provide clear ways to combine content across the framework. A content repository is the name given to a database of media items such as articles, videos, music files. Ontop of the database, a content repository implements several ways of accessing and organising content so that it's easy to find and navigate. TagMyBits differs from existing repositories by allowing users to store and navigate content via the explicit relationships between items. These relationships can be contextual and are defined when data is added to the repository.

Objectives

• The ability to add new types of content to the framework without having to change the storage model.
• Allow a highly ?exible organisation of content.
• Allow content to appear at various points in the framework so that popular clips and segments don't need to be manually copied.
• Display content associated with what's currently being viewed.

Bits Tags and Chains

All of the data in the framework can be described as either a Bit, a Tag or a Chain. Bits contain the content of the framework and are used to store content data; Tags allow users to categorise data by assigning bits key words or phrases; and Chains describe how bits are associated with one another.

Chains provide the flexiblity of the data model by allowing users to connect content together any way they see fit. Bits can exist in zero or more chains making it possible to reuse items of content across the framework in different contexts. In the diagram opposite B₂ has been used in both chain C₂ and C₃. Further, consider the case where B₂ is a picture of a frog, B₃ a toad and B₄ a fish. In this instance it is possible to imagine a context where these bits are and are not related. At C₂ a user may only be interested in images of amphibians, hence they see the pictures of the frog and the related picture of a toad. At C₃ the user may be interested in images of pond-life or animals hence they see the frog, toad and fish images.

Dynamic Types

Since all data in the framework is stored as a bit, the system needs a method of resolving the different types of content. This is achieved by using a system tag associated with the bit to define its type. Bits containing scripts are used to render the different types of content. This means that new content types can be added on-the-fly and is as simple as creating a new bit in the framework. The handler for a type is identified by a system tag matching that of the content being rendered.

Going Further

While accurately targeting recommendations is a challenging task, it becomes even more di?cult when there is a lack of historical data, such as in the case of new or unpopular items By storing the relationships between items in the framework it is possible to strengthen existing search and recommender systems to include information about the structure and context of content. This can be used to provide more relevant results or initial data when statistical reports on user behaviour are not available.

To read more about the project, please download the full project report above where the source code for the project has also been made available.