WO2006008710A1 - A method for refining media programming - Google Patents

Abstract

A method is described for refining media programming content comprising entertainment programming, educational applications, or the like. A content provider sends an application to a user device. At the user device the application is executed and feedback from the user is obtained and stored. This feedback is sent back to the content provider which then analyses the feedback and automatically adapts the application according to the results of the analysis.

Description

DESCRIPTION

A METHOD FOR REFINING MEDIA PROGRAMMING

The present invention relates to a method for refining media programming content within a system of devices and also to an apparatus therefor.

The strength of traditional one-way broadcast (transmitters to radio/TV) is the potential for huge coverage over a known geographic area. Cost to the broadcaster is relatively low per listener/viewer. Cost to the service provider can be low for radio, higher for TV production. Cost to the audience is low for radio, higher for TV (though it becomes cheaper with shared access). Routes to gather user responses ('the back-channel') are via phone-ins, or SMS/email are not very expensive, but the cost may inhibit participation by poorer audiences. In general, the broadcast systems currently deployed only elicit single answers or votes on an issue. Furthermore using broadcast there is no means to deliver personalised feedback to users on their responses or segmenting or grouping of users according to user wishes. It is known that networks, for example the internet, allow PC users to communicate over very large geographic areas. Application or content providers are able to receive user feedback. In one example relating to systems using Microsoft Windows™, a user can report system and program errors to Microsoft. This reporting system allows Microsoft to track and address operating system, Windows component, and program errors. To report the problem, technical information about the problem is collected and then sent to Microsoft over the Internet. The information Microsoft receives is used by programming groups at Microsoft. If a similar problem has been reported by other people, and more information is available, a link to that information will be provided. The scheme focuses on fixing the reported errors and/or suggesting work-arounds; representing a personalised feedback to users on their responses. An inference is that reported problems are dealt with by personnel at Microsoft, such as working on solutions to errors which are frequently reported. This is a typical example of measures undertaken by software vendors in respect of adaptation of their software for error rectification.

It is an object of the present invention to improve over the prior art.

According to a first aspect there is provided a method for refining media programming content within a system of devices, the method comprising: - at a first device: o sending an application to a second device; at the second device: o receiving the application; o executing the application; o obtaining and storing user feedback; said feedback obtained during execution of the application; and o sending the user feedback to the first device; and at the first device : o receiving the user feedback; o analysing the received user feedback; and o automatically adapting the application according to the results of the analysis.

Advantageously, an application is automatically adapted according to the results of the analysis of user feedback thereby minimising on-going resource requirement for application editing and/or compilation by the content/service provider. Furthermore, adaptation of the application according to an individual user's feedback can be automatically provided.

Iterative refinement of the media programming is provided to a community of users using analyses of their responses to programmed elicitations/queries and/or using their automatically-tracked behaviour when playing the programming. Analysis can be greatly speeded when possible by encouraging the user to use specific responses, such as certain key-presses or words from a limited vocabulary.

Advantageously, responses can be elicited from the audience on a wide range of topics from local issues such as jobs, crime, water supply.... all the way through to critiques of politics, sports events and latest fashion in humour. The tools provided to undertake analysis of feedback enable the creation of new programming quite readily, for example for distribution as a follow on to a competition or request for comment.

Templates for different application types can assist with application creation, setting styles or genres of interaction with which the user becomes familiar, but also preparing styles of the analysis of subsequent feedback.

Selection (or creation) of appropriately structured audio assets (prompts/queries) used in content assembly increases the likelihood of responses that can be analysed easily at a later stage. Selection of assets can be aided by appropriate mark-up (description) of the assets.

Authoring of an application may automatically create a feedback analysis tool (response structures and associated analysis mechanisms), along with the program's context and knowledge domain/scope to enable easy assimilation of responses. One of the results of feedback analysis to a particular program is the potential to create further programs that meet an identified need of the audience, or identify subgroups of the audience with specific interests or needs. For example, the number of users of the application that respond to the application in a particular manner, those users which belong to a particular answer range, or those users which only partially use content items. These subgroups can then be addressed more specifically. As an example, a high incidence of incorrect user responses may indicate that a query was not well worded. This might similarly be indicated by a large number of incomplete playback operations (where users "barged-in" and did not listen to the complete question). If sufficient responses to this effect are noted during feedback analysis, then a replacement item can be created and distributed to replace the problematic one. If queries or prompts are marked up in a regular way with their characteristics, replacing one item with another can be done very quickly by searching out the best match item with the same characteristics. It may be the case that many responses are received that show failings of the original application, causing confusion, and an update to the application needs to be distributed which clarifies the topic in question. The responses from the target audience may show differences in interests/opinions, and therefore in the need for further media programming. A single distribution list may evolve into more focused subgroup distribution lists. Furthermore, there may be an indication from the responses (or lack of responses) from some members that the program is irrelevant for them. In such cases it will make sense to modify the distribution of subsequent follow- on programs to exclude the uninterested parties.

When an application is a compilation of material, for example an audio newspaper, this material may come from different topic providers. Each of these providers may work with a subtemplate to contribute their part to the automatic compilation, the use of subtemplates therefore simplifies their job - lower skills and computing resources are required by each contributor to the whole.

Content templates may also inform (or 'template') the likely style of feedback analyses required. For example, an 'interactive radio' template may not track the user's playback behaviour, only gather explicit responses. In that case implicit user behaviour may not be analysed. Alternatively, a census application template may set up a data-mining tool to analyse the statistics that will be gathered. Preferably, the analysis results are converted into XML-based instructions for the content authoring tool to execute and thereby adapt the application.

The ability to exploit audience feedback in order to refine the content that is broadcast, or indeed create and refine distribution lists, is powerful because it enables increasingly relevant content to be created, and increasingly accurate targeting of that content to specific audiences or individuals, hence increasing the engagement of the audience. The functionality described here can be implemented in software, hardware or a combination of these. The invention can be implemented by means of hardware comprising several distinct elements, and by means of a suitably programmed processing platform. Accordingly, another aspect of the invention provides an apparatus comprising processing means operable to: send an application to a second device; receive from the second device user feedback on the application; analyse the received user feedback; and automatically adapt the application according to the results of the analysis.

The instructions may be stored on an electronic memory device, hard disk, optical disk or other machine-readable storage medium. The instructions may be downloaded directly to a device via a network connection, or by means of removable media.

Further features and advantages of the present invention will become apparent from reading of the following description of preferred embodiments of the present invention, given by way of example only, and with reference to the accompanying drawings, in which:- Figure 1 shows a system level view of elements of the invention and the messages passing between the elements;

Figure 2 shows templates and sub-templates for an application;

Figure 3 shows various degrees of topic refinement making up an application; Figure 4 shows automated iterative refinement of content programs and distribution lists;

Figure 5 shows a schematic of the sequence of preparing interactive, non-linear, two-way audio applications;

Figure 6 shows a system level view of the authoring tool and interactions between its primary components;

Figure 7 shows components and processes of the central 'tuning system'; Figure 8 shows a Feedback Analysis Tool (FAT) outputting a report file according to the Content Provider's (CP) specifications in spreadsheet format;

Figure 9 shows applications published to Regional Content Centres (RCC) where editorial control and distribution control is maintained; Figure 10 shows a first example embodiment of the State Health

Program; and

Figure 11 shows a second example embodiment of the State Health Program.

A Voices in Your Hand (ViYH) system is used in the following to illustrate the present invention. The system has an infrastructure which is focused upon provision of content and personal communications in an asynchronous manner and includes mechanisms to describe and distribute programming to ViYH users. This programming is primarily aimed at the provision of speech-audio (for social / informational / educational / entertainment purposes) as well as provision of music-audio, but applications can also include textual and graphical content.

Furthermore, the applications are structured with mechanisms to elicit responses (such as keypresses) from the user in order to navigate through the application, as well as elicit keypress or speech responses to spoken queries. Therefore, immediate local interactivity is possible on a handset, as well as asynchronous two-way exchanges between users and service providers.

One particular example is the successive refinement of health and medical information distributed to community health workers (CHWs). Community Health Workers need relevant training material based on their unique and personal levels of understanding of a given topic (e.g. Immunization Procedures.). CP/SP's (Content Providers / Service Providers) at the State Health Program prepare a set of system-based Immunization Procedure training lessons that are designed to evaluate the pre-existing knowledge of users. The queries embedded in the learning applications assess CHWs Immunization Procedure expertise. The CAT/FAT (automatically) prepares subsequent training lessons based on user responses to previous lessons.

Now an example embodiment illustrating various aspects of the invention is considered. The Content Authoring Tool (CAT) and the Feedback Analysis Tool (FAT) are at the centre of automated iterative refinement of content programs and distribution lists in the present system as shown in

Figure 4.

Service providers (SP) are specialists in one domain, and are able to spend time if needed working with individual users on issues such as health, employment or particular entertainment or lifestyle interests. Content providers

(CP) are experienced in formulating applications with appropriate levels and mixtures of informational, educational and entertainment material to keep users motivated to use and benefit from the system. Possibly taking input from a SP, a CP will use a CAT to produce an application, creating a FAT set-up in the process.

Referring to Figure 9, applications are published to Regional Content Centres (RCC) where editorial control and distribution control is maintained. The RCC maintains databases of users and their user-profiles, and individual subscriptions to distribution lists. Applications are distributed to a user if their profile matches the intended target audiences for that application, or if they have subscribed to it.

Users dock and refresh their devices at Internet Tele-Centres (ITC) or refresh their data-cards using a Data-Card Refresh Service (DCRS). At refresh time, the applications identified for a user are downloaded to their device (after the upload of voice messages ('vmails') and responses to earlier applications they have played). Responses to the new set of applications are uploaded during the next dock and refresh cycle, and directed to the specified SP.

The embodiment described here articulates the core functions and features of the CP/SP to CAT to User to FAT to CP/SP feedback loop or 'Tuning System', as illustrated in Figure 4. It also details the interactions between system components and provides details of use for Community

Health Services. Figure 1 is a system level view of elements of the invention and the messages passing between the elements.

Figure 5 illustrates schematically the sequence of preparing interactive, non-linear, two-way audio applications for the present system. The final application program consists of program objects, audio and textual assets.

Figure 6 is a system level view of the authoring tool and interactions between its primary components

Figure 7 illustrates components and processes of the central 'tuning system'. The CAT facilitates the assemblage of content programs for distribution to users. A number of CAT features also directly support the iterative refinement and tuning process.

The CAT's role in the Tuning System is now described:

1. Asset, Query, and Program marking a. The CAT enables the 'Structuring and Markup of Resources.' b. Audio assets, mark-up language queries, and individual content programs may be labelled and categorized in order to facilitate efficient automated access and assemblage of resources;

2. Program Architecture a. The CAT creates the formal program structure of each content program. This program structure provides a schema for later analysis by the FAT. The CAT also produces additional information on appropriate tools, their parameters, contextual or domain information to assist in subsequent analysis, referenced to the same schema;

3. Distribution a. The CAT can specify both the structure and form of content, the recipients of content (users) for later distribution by the RCC, and the recipients of content responses for analysis in the FAT. b. The CAT can receive instructions on distribution (program recipients) and content structure from either the CP or the FAT. c. The CAT can automatically implement XML based distribution list and/or content assembly instructions that are received from the FAT;

4. Editorial Control (optional) a. The CP can play an editorial review role in ensuring appropriate distribution lists and content structures are prepared for output. b. The CAT provides the CP with direct editorial control of content programs and recipients of content programs. c. The FAT allows the CP to be included to moderate the automated process by manually checking both the automated

FAT instructions and the CAT implementation of those instructions, before output of an adapted program for the next iteration;

5. Authoring a. The CAT is the Authoring Tool for the assemblage of content resources into program structures. b. The CAT automatically assembles content programs using the library of marked-up assets (shared resources) as instructed by the FAT. The FAT gathers program queries and responses into a discrete dataset and performs various analysis functions that both make inferences on the responses to program content and the users as a data-mining tool, but also prescribe the iterative refinement of subsequent content programs or distribution lists. Referring to Figure 8, the FAT outputs a report file according to the CP's specifications in spreadsheet format. In addition to the items specified by the CP, the spreadsheet report may for example also include:

^■ Program name

■ Time stamp ■ Summary report information

^■ Total number of questions included in the report

^■ Total number of respondents included in the report ^■ Total number of responses included in the report

For MS Excel output, audio responses can be linked to the original audio file using a URL for the location of the file, enabling the SP to launch and listen to the audio file. Using MS Excel or another spreadsheet tool, the CP/SP analyses the report according to their specific needs. The flexibility and functionality of a spreadsheet program deepens the CP/SPs understanding of the analyses of application responses. For example, they can create charts or graphs of application responses and the demographics of respondents. This makes extracting the trends, opinions, and perspectives contained in the data more accessible.

The following embodiments illustrate practical applications of CAT/FAT tuning mechanisms in one context- that of community based Health Services. The focus here is on leveraging the system to improve delivery of relevant training information based on the needs and priorities of Community Health Workers as users. Alternative embodiments are applications for subscribers to an interactive community radio journal, community job board, or local entertainment programs. Two central tuning features are demonstrated: 1. Customized Distribution Lists 2. Adaptive Content Authoring

These scenarios and the key features of the CAT/FAT system must be understood in the context of the present system and its focus on the needs of users at the bottom of the pyramid.

In an example embodiment the State Health Program uses the present system for preparation and distribution of training content to locally based Community Health Workers (CHWs). The Community Health Workers need relevant training material based on the local priorities and needs. CP/SP's at the State Health program prepare an exploratory polling application in the CAT for distribution to CHWs throughout the state. The polling application queries CHWs on the training priorities for their region. For example this may include Aids, children's health, immunization procedures, sanitation, and conducting home visits. The FAT receives the polling results of this survey. Based on the responses, customized distribution lists for training content are automatically prepared and sent to the CAT for execution. State CP/SPs can maintain editorial control of the process in order to ensure that content distribution aligns with regional policies and procedures. This embodiment is now described with reference to Figure 10.

1. Polling Application Authoring and Criteria Setting a. The State CP/SPs author a polling application in the CAT specifically designed to extract health content distribution priorities based on local needs. A library of such polling applications may be created for ongoing evaluation of content distribution practices and their effectiveness. b. The State CP/SPs also establish response analysis policies in the FAT. These policies establish the criteria for customized distribution list creation. E.g. those requesting Aids training receive lessoni: Aids from the content library. These policies may increase in sophistication. For example, the State CP/SPs might require that those requesting Aids training first complete the entire set of lessons on Sexually Transmitted Diseases.

2. Polling Application Distribution a. The polling application is distributed to local SPs. Local SPs complete the polling application and send the results back to the State CP/SP for analysis.

3. Polling Results Analysis a. The structure and format of polling application queries are extracted from the shared library of resources (content library).

Poll responses are received from users via the RCC interface to the ViYH system and its end-users. The FAT analyses the poll results by integrating information from these two sources. b. The FAT performs analysis functions based on policies set by the CP/SP in step 1.

4. Distribution Instruction Preparation a. The FAT converts the analysis results into XML-based instructions for the CAT to execute. These instructions provide the CAT with the form and structure of customized distribution lists. b. To support editorial control, the CP/SP is informed (e.g. via a popup screen upon opening the application for instance) that customized distribution list instructions are ready for execution by the CAT. c. Instructions Conceptual Example: Send users #1002, #1003, and #1005 Aids Health Lesson 1 , Send users #1004 and #1006

Immunization Procedures Lesson 1. 5. CP/SP Editorial Control a. The CP/SP may optionally edit the CAT instructions prior to execution. 6. Automated Distribution List Creation a. The CAT creates a distribution list based on FAT instructions. 7. 'Tuned' Content Distribution a. New content is distributed to users based on the 'tuned' distribution list. (As examples, further training material on may sent to those users who failed to pass a test, or more specialised material on a specific health concern sent to those users who indicated that concern as being high on their local priorities)

In a further embodiment the State Health Program seeks to customize training content based on the levels of pre-existing knowledge of users

(CHWs). The focus here will be on CHWs knowledge of Immunization

Procedures. Community Health Workers need relevant training material based on their unique and personal levels of understanding of a given topic. CP/SPs at the State Health Program prepare a set of system based Immunization Procedures training lessons that are designed to evaluate the pre-existing knowledge of users. The queries embedded in the learning applications assess CHWs Immunization Procedure expertise. The CAT/FAT automatically prepares subsequent training lessons based on user responses to previous lessons. This embodiment is now described with reference to Figure 11. 1. Asset Mark-Up and Cataloguing a. In order to prepare customized content the CAT draws from a resource of catalogued training assets (objects) focused on the particular content area (Health in general and Immunization Procedures specifically in this case) b. The CP/SP marks up and catalogues these assets or programs prior to engaging the Adaptive Content Authoring process. 2. FAT Criteria Setting a. The CP/SP establishes response analysis policies for the FAT. These policies establish the criteria for customized content creation in the CAT. b. Policy Example: If users score under 50% on the lesson 1 final exam, they must repeat lesson 1. Users 50-75% must take a review lesson, and users who score 90-100% may move on to lesson 2.

3. Training Application Distribution a. The learning application is distributed to users. Users complete the training application and send the results back to the CP/SP for analysis.

4. Training Application Analysis a. The structure and format of training application queries are extracted from the CAT. Application responses are received from users via the RCC. The FAT analyses the training application results by integrating information from these two sources. b. The FAT performs analysis functions based on policies set by the CP/SP in step 2

5. Content Authoring Instructions Preparation a. The FAT converts the analysis results into XML based instructions for the CAT to execute. These instructions provide the CAT with the form and structure of customized content. The CP/SP is informed that customized distribution list instructions are ready for execution by the CAT. b. Conceptual Example: Send user #1002 basic lesson 2, using assets a, b, and d. Send user #1004 advanced lesson 2 using assets, f, g, and i.

6. CP/SP Editorial Control (optional) a. The CP/SP may manually edit the CAT instructions automatically-proposed by the FAT prior to their execution.

7. Automated Content Authoring a. Upon CP/SP approval the CAT automatically authors the next training program based on the FAT's instructions.

8. Tuned' Content Distribution a. The Customized Content is distributed to users. In a preferred embodiment a schema of elements and attributes are used to describe the media objects used in the creation of an application, as well as the responses received from end users. Schema present relevant attributes of the content assets to enable rapid selection of candidate assets when building a content application.

The schema for each asset specifies the makeup of response(s) that the user provides to it. This schema then contributes to application authoring. Furthermore, when the CAT builds an asset into an application, the schema for responses to that asset are also built into the feedback analysis tool, for later analysis of responses to that asset by the service provider.

Using the schema like a grammar to specify topic area and relationship with other content assets further speeds the collation of materials and creation of a new application prompted by user responses. These schema can be used to build templates and data-analysis (data-mining) or natural language analysis tools for a more intelligent subsequent analysis of the users' feedback. Referring to Figure 2, templates for different application types assist with application creation, setting styles or genres of interaction with which the user becomes familiar. Templates also assist with preparing styles of the analysis of subsequent feedback. Selection (or creation) of appropriately structured audio assets (prompts/queries) used in content assembly increases the likelihood of responses that can be analysed easily at a later stage. Selection of assets is aided by appropriate mark-up (description) of the assets. Templates are available for different application genres, for example:

^■ Interactive Radio

^■ Non-linear education

^■ Branching stories

^■ Census These templates define the structure of interactions with the user (the series of prompts, content, navigation options, expected responses etc) and also detail the make-up of each response possible to each object (e.g. a number between 1 and 5, a spoken description, a choice of three options). Moreover, the templates set up the automatic tracking of what the user plays or does not play in the application and its subsequent reporting back to a CP or SP. Each template can be expanded and edited as needed, but the essence of the application-type (genre) remains. In this way, users become familiar with a certain style of application, interaction and privacy -making a series of interactive applications easy to use. The content author has access to a large number of audio assets - informational, educational, entertainment - as well as query objects related to those assets. In order to assist selection during the application authoring process, in a preferred embodiment assets are grouped according to categories (e.g. health subdomains), or style (e.g. open-ended or closed question). The content provider also aids the subsequent response-analysis task by marking objects in the application with the appropriate analytic tools or parameter settings that should be used on the user responses. In addition, at content authoring time, the provider sets up 'user stereotypes' or answer ranges to aid in the categorization of user responses. These ranges may for example identify people inferred to be at risk for a certain disease.

When applications are created, query objects identify the parameters of the responses that can be acquired from the user. The CAT maps these parameters and their unique application-object identifiers to the FAT for analysing the responses to that application.

Query objects can demand several types of response from the user:

■ Choice of options(s) from a list ^■ Spoken response

^■ Open-ended response

In a preferred embodiment, through the control of 'history-tracking' parameters, the user's path through a branching application or the number of times they play a particular object in the media application is traced, This implicit information also forms data that can be automatically forwarded to the

SP for analysis along with any explicit user feedback.

The use of identifiers for applications, query objects, users, and user- responses enables each user-response to be correctly located within the analysis structure of the feedback analysis tool. In this way, it makes it easy for the SP to either follow a single user's path and responses through the application, see which paths were or were not exercised by the audience, or alternatively look at all the users' reactions to a particular question in the application.

Some topic areas or interaction invitations may be more effective than others in engaging the audience in the subject domain, whether it be health advice, news or musical entertainment.

In an example embodiment a first application exposes a certain area of content of particular relevance that can then be followed up by a second application exploring that topic in more depth. This process can proceed automatically or be under editorial control, for example as a result of user- votes on different topic areas or the popularity of certain application objects for user playback.

Referring to Figure 3, varying degrees of refinement can be made to the content broadcast by the present system. At the highest level is the creation of a new channel: addressing issues and providing information to a segment of a poor community that is not achieved elsewhere by other means. In an example embodiment material on a range of topics is organised hierarchically. The top level is included in a first application broadcast. Then, according to the measured popularity or interest aroused by different areas in the application, a follow-up application broadcast including the next level of detail on those topics is broadcast.

In the case that the application consists of alternate branches, or paths that the user can navigate, the audience feedback may indicate that a whole branch is effective or not effective, so branches can be grown or deleted in the next iteration. In an example embodiment such 'application refinement' is automated by counting the number of users who play an item all the way through, or play it back on several occasions. If the number of positive responses exceeds a threshold, then the creation of a programme that elaborates upon that item is automatic. The responses from the target audience may show divisions in the interests/opinions, and divisions in the needs for further programming. In an example embodiment the "learning system" creates new distribution lists necessary for subsequent, more specialised follow-up programming which will be relevant to distinct subgroups of the audience. In this way, a single distribution list evolves into more focused subgroup distribution lists. Identification of subgroups can provide the impetus for targeted "mailshots", visits or telephone calls. Furthermore, there may be an indication from the responses (or lack of responses) from some members that the program is irrelevant for them. In such cases, the distribution of subsequent follow-on programs is modified to exclude the uninterested parties. In both of these cases, the CAT-FAT combination will have an impact upon distribution lists (and profiles) stored centrally at the RCC (regional content centre).

In the description above, and with reference to the Figures, a method is described for refining media programming content comprising entertainment programming, educational applications, or the like. A content provider sends an application to a user device. At the user device the application is executed and feedback from the user is obtained and stored. This feedback is sent back to the content provider which then analyses the feedback and automatically adapts the application according to the results of the analysis.

Claims

1. A method for refining media programming content within a system of devices, the method comprising: - at a first device: o sending an application to a second device; at the second device: o receiving the application; o executing the application; o obtaining and storing user feedback; said feedback obtained during execution of the application; and o sending the user feedback to the first device; and at the first device : o receiving the user feedback; o analysing the received user feedback; and o automatically adapting the application according to the results of the analysis.

2. The method of claim 1, wherein for a successive user feedback the application sent to the second device comprises the automatically adapted application from a previous iteration of the method, the sent application in turn being automatically adapted according to the results of the analysis of the successive user feedback.

3. The method of claim 1 , wherein the step of obtaining and storing user feedback further comprises: requesting specific responses from the user; receiving and storing said specific responses.

4. The method of claim 1 , wherein the step of obtaining and storing user feedback further comprises automatically tracking and storing information on the usage of the application.

5. The method of claim 1 , wherein the step of analysing the feedback comprises counting the number of users of the application that respond to the application in a particular manner.

6. The method of claim 1 , wherein the step of analysing the feedback comprises categorising the feedback into particular answer ranges.

7. The method of claim 1 , wherein the step of analysing the feedback comprises counting the number of users using content items incompletely.

8. The method of any of claims 5 to 7, wherein the step of analysing creates XML-based instructions.

9. The method of claim 1 , wherein the step of adapting the application comprises automatically replacing one content item with another content item having the same characteristics.

10. The method of claim 1 , wherein the step of adapting the application comprises automatically adding further programming on a topic relevant to the user.

11. The method of claim 7, wherein the step of adapting the application is by means of executing XML-based instructions.

12. An apparatus comprising processing means operable to: - send an application to a second device; receive from the second device user feedback on the application; analyse the received user feedback; and automatically adapt the application according to the results of the analysis.