Official Website of MyMedia Collaborative Project

The facts about MyMedia Recomender System initiated by BBC

In earlier days, when the Radio was utilized as the element of entertainment the listeners were used to habituate the programs that were played by the channel administrators. Later on when the tailored or customized channels came into the reality, the users had the options to select their programs as per their choice. They can compare the programs along with its attractiveness. After that when the Television channels came into the realism users were had the option to select their selections as per the audio and video quality of the channels.

In the early 80's and 90's the private television and radio channels came with their fullest and several options of entertainment programs that user can select as per their choice. But, it was not the exact that the channel marketers were looking for. Because, there are several private channels that are attracting the viewers and listeners with their most fashionable and colorful programs. Due to the lack of information in proper time the users were not able to hit the programs in proper time. That affected the channels a lot. After that the professionals came and they were fully engaged to build up the viewer's community through various data analysis and they implemented several surveys to gather data from the viewers through several promotional programs. The term `TRP' came in to the scene. The "Television Rating Program" or TRP is the parameters that is helpful to the tele-analyst to get the exact viewer traffic to their channels.

In these days, users can catch up TV and radio channels through their mobile, Tab, PC and from any android software based devices. Now, this is really a fix to the data analyst to track the users and their selections. To entrust on the quality of their broadcast and presentation the satellite based channels are propagate their programs to the viewers and users with the trust that people will view their channels as they are channelizing good and entertaining programs. But, is it really effective?

To catch up the real picture of the market the BBC, the largest Air and Web based television and radio service provider has conducted a field survey that they named with 'MyMedia'. The complete name of the project is - Dynamic Personalization Of Multimedia. The survey report was finally and officially placed before the public on 2010. Chris Newell, the lead Technologist of BC was the man who was entrusted by the media group to conduct the survey project associated by Bart Knijnenburg, Head of Computing at Clemson University.

Basically the entire field survey was the framework for preparing the master algorithm for an open source recommender system software framework. In the executive summary of the final copy of the project, Chris has proclaimed that the BBC MyMedia field survey has targeted the users those are using web based TV and Radio service catch-up services through MyMedia recommender system software. Those field trials establish the performance of the recommender system that is based on explicit, implicit and non-personalized feedback by the users for a certain period of time.

While implementing the field study, the BBC (The British Broadcasting Corporation) has targeted the users those are using web TV and Radio catch-up services for free of cost for 7 days or more after it has broadcast in UK. This service was limited to the users of UK for absolutely free of cost and the entire cost was given by the BBC out of its license fees.

The BBC offers 10 television stations and 57 radio services which collectively broadcast about 500 programs every day. Many of these programs become accessible in the catch-up service right after broadcast. The catch-up service allures several million exclusive users per week and each customer watches an average of 3-4 programs weekly.

The variety of programs offered by the service is quite broad. Some material, just like news programs, becomes less interesting to users gradually. Other material has a long-term potential life-time and many programs may be prepared into series that are best seen in the correct order. These aspects should be regarded when making suggestions. The present viewers is generally young, male and acquainted with the Internet but it is modifying as the service turns into better identified and reaches a broader viewers.

Presently, many people are utilizing the catch-up service to enjoy programs they have skipped or have overlooked to record when they were actually broadcast. In such cases they know precisely which program they are searching for. The existing user interface utilizes a channel and plan based strategy and is well matched to this kind of action. However, the user program is less suitable for the users who are searching through the catalog searching for something to look at with no specific program in mind. Because of the big size of the catalog this procedure can be very time-taking and tedious. Simple suggestions are applied in portions of the current interface, but they're a minimal section of the interface. A new user interface that is more suggestions-based, instead of schedule based, will be of significant benefit in this situation.

Some determination to the Recommender System

The inspiration to the BBC MyMedia Recommender system is to improve utilization and enhance user fulfillment in the searching mode of intake where the user doesn't need a clear concept what they may be searching for. The size of the catalog is probably to boost considerably in future years as it becomes possible to increase the duration of the accessibility window and content may also become accessible from the BBC's significant archives. To offer a sufficient user experience when looking a larger collection a new type of user interface might be expected.

A second inspiration to the BBC Recommender system will be the probability of a more customized service to the website in general. A higher degree of strength is instructed at marketing activities in most BBC services. The website offers the probability to immediately deliver targeted offers using the recommender. One area MyMedia investigated was the incorporation of 'online upgrade' techniques into systems that need offline coaching. This enables innovative collaborative selection systems to support new users and products devoid of having to await another time-taking training cycle to be finished. We lately ran an inner field trial to find out if these online techniques were efficient and how good they may scale. The outcomes were very appealing, defeating our best meta-data-based method regarding the recognized quality, range and uniqueness of the suggestions.

We are also discovering fresh user interfaces and the probability of offering recommender systems as a separate application, instead of an integrated element of current services. These connections try to give the user more manage over the referrals and provide higher transparency. While in a beginning stage it is becoming clear that offering this type of control absolutely minimizes user aggravation. Moreover, the further feedback that is gathered can produce a important improvement to the suggestions. Preferably, it can be easy to make some of these prototypes openly offered later on, to find out whether a stand-alone recommender application will be efficient and valued.

User Selecting

60 field trial individuals were enrolled by a market research organization and supplied a monetary incentive to finish all components of the field test. The members were enrolled based on the following shape:

Individually own and consistently use a Laptop or PC
Cozy using a computer key pad and mouse unaided
Never involve any professional equipment to utilize a computer
Offer an internet network at home
Are usually not "contra-web"
Will not be internet specialists or beginners
Use the present catch-up service consistently

This user profile was designed with example of the person as defined in deliverable and to make sure sufficient discussion with the entire field trial user interface. The members were enrolled in equivalent numbers from 5 areas across the UK and in all of these areas the individuals were well balanced based on the following standards:

Similarly separated by gender
Likewise divided into 3 age brackets (18-35 years, 36-50 years and 51-66 years)
Both equally divided among heavy users (3 times weekly or much more) and mild users (1-2 times weekly) with the catch-up service.

For recommender type training the field trial also applied unknown data recorded from customers of the BBC's active catch-up service and participants of a web-dependent audience analysis panel. This assured there was a significant quantity of explicit and implicit feedback accessible to the recommender system during the trial regardless of the reality that the range of direct members in the field trial was comparatively small.

Contributors were recognized and supervised using the BBC's user identification system that is based on consistent browser cookies. This needed the user to sign-up on the beginning of the trial and to sign in once on each computer they used afterwards. This strategy permitted individuals to work with the system both in the home and at the job whilst keeping the same identification, as defined in presentable manner.

Data Evaluation Methods

Quantitative analysis: Structural Equation Models (SEMs) were designed depending on the survey data, recorded data, and the circumstances manipulated in the test. We first ran an exploratory aspect analysis model on the forms to ensure that the constructs we expected to measure using the questionnaires indeed form conceptually unique constructs. Then we run a confirmatory model on the uncovered constructs, and generally link them to one another and to the situations and signed data in a regression path model.

Qualitative research: The members' reactions were transcribed and evaluated using based theory research. Appropriate passages were classified, and schemas designed to depict conceptual interaction produced from these paragraphs (e.g. "recommendation excellence" effects "choice fulfillment"). These communications represent the outcomes of the analysis, as they offer proof for the interaction between ideas in the assessment framework.

Field Trial Assessment:

Assessment System: Work Program helped the BBC MyMedia field trial with the growth, application and testing of assessment metrics. A general user-centered assessment framework was designed and has become verified in several managed experiments along with the field trials themselves. The framework advised each qualitative and quantitative evaluation of field trial details. Beside the natural value of the composition, the ultimate goal is to determine generic human-recommender connections and user experience guidelines, in addition to examine the particular field trial techniques.

The composition ensures that outcomes from numerous trials could be compared and incorporated into a solid popular comprehension. The selection of employed methods offers a total knowledge in recommenders which is both wide and profound. The qualitative assessment provides a very intricate analysis of personal users' ideas of the field trial techniques and the impact of a recommender system for the user knowledge. Quantitative assessment, on the contrary, enables us to back our theories with uncertain mathematical facts based on data from a bigger number of customers. Managed studies allow us to precisely figure out the impact of specific recommender system factors. Field information, however, can be utilized for monitoring extended-term effects in environmentally valid settings

User-centered Assessment Framework: The framework could be used to assess any recommender system, attempts to describe the impact of the purpose system factors on the users' behavior. The structure is user-centered, because this makes the link by clearly calculating the mediation through subjective system elements (how the user interprets the program) and the user expertise (how the user thinks the connections using the system). Moreover, it puts this description into framework by taking the features of the person and the scenario into account. The structure takes the objective system features, and decides their immediate impact on the users' belief of the system. These thoughts or subjective system factors include functionality, excellence and benefit the system. Note that while these subjective system factors are subjectively calculated, they don't say anything about the users independently: the concentration is on the process and how a user interprets it.