In order to evaluate the system a user interface has been created to allow users to use the services that are offered by the EPG. This allows users to create their own accounts using the same name that they use for accessing Facebook. The prototype EPG has three areas to display the recommended lists of programmes based on the three recommendation techniques. Moreover, it contains ON and OFF buttons to turn on and turn off the EPG. When the user turns the EPG on, the recommended lists of programmes will be displayed. By choosing any one of the recommended programmes, details of that programme will be displayed which include the programme’s description, genre, length and production year. Additionally, the address of the selected programme is added to the displayed details in order to check the original online source.
Another display area has been added to the interface for doing a direct keyword search. Two buttons have been added for this purpose; the first one is used to input the search keyword, where this keyword will be sent to the context-aware module program and from
147
there the search agent leader will search the online content sources. Thereafter, by pressing the second button, the system will display the list of programmes that are recommended. Figure 5.31 shows a snapshot of the designed EPG.
The interface also allows users to watch the selected programmes, where a button has been located in the interface for this purpose. This button opens an Internet browser with the address of the selected programme. The system then automatically sends the programme’s features matrix of this watched programme to the context-aware module so that the user viewing history can be updated.
In order to enhance the recommendation process and avoid the cold-start problem, the user interface has been provided with a list of categories. The user is promoted to select their preferred categories during the registration stage. This helps the system to provide a list of recommended programmes based on the selected categories (Collaborative Filtering Technique) even when the user hasn’t watched any programme and therefore has no viewing history. Furthermore, the users can add additional categories at any time when they turn on the EPG.
Because some of the online content sources are pre-paid or subscription sources, the user interface has been provided with an option to either includes the pre-paid sources and sources that need subscription or not. This is done through providing a choice within the designed interface and the user can select ‘Yes’ to include those sources or select ‘NO’ to avoid the search results that come from those sources during the registration stage. The full configuration software of the designed EPG is located in Appendix VI which combines all the functions within the proposed system.
148 F ig . 5.31 S na pshot of th e De sig n ed EPG
149
5.7 Summary
In this chapter, the implementation of the proposed EPG system has been described. The different functional blocks have been created using PHP and Matlab. According to the current implementation, the EPG system has the following features:
1. The content search module is able to retrieve the programmes’ metadata from all the connected online video content sources which are YouTube, BBC iPlayer, Dailymotion and The Movie Database.
2. All of the metadata formats that are retrieved from the different online content sources are converted to a unified format, which include 21 programme features, that is used within the remainder of the system.
3. Multiple recommendation schemes have been used to enhance the recommendation process which are content-based, collaborative filtering and context-awareness. 4. Integration of social networks has been realised through an interface to Facebook
from which a list of programmes is obtained based on the users’ preferences and their friends and is used to enhance the recommendation operation.
5. The system is generic and flexible to any expansion in the number of online video content sources where the core system architecture remains the same as new online content sources are added. The only bespoke module is the search module which has to be unique for each new online content source.
6. The system is provided as a cloud based service in which all recommendation operations are carried out in a centralised server with the results being stored in that server and made available to the user when they log onto the EPG.
7. A filtering stage has been created to filter the recommended programmes based on the resolution of user viewing device and network bandwidth.
The next chapter will discuss the evaluation of the EPG functions based on the implementation which has been described in this chapter.
150
CHAPTER SIX
151
Chapter Six
6.1 Introduction
The previous chapter illustrated the details of the EPG system implementation. This chapter describes the processes through which the system was tested and evaluated. In order to validate the proposed architecture, the EPG system has been evaluated through assessing its performance according to the main functions that were presented in chapter three. As mentioned in the previous chapter, in order to evaluate the proposed system, a basic user interface was designed to allow the users use the system and its services.