Adaptive content presentation

9 minute read

Updated: December 07, 2020

Part 1 – Static approaches
Part 2 – Dynamic approaches
Summary
What to show and how to show is called adaptive content presentation
- Static approach – content is already there. How do we show them on the screen
- Dynamic approach – need to automatically decide what to show to the user
  - Complicated, but this is where AI comes in

Part 1 – Static approaches

Adaptive content presentation – two steps

Content adaption – how to select most appropriate content for the current user in the current situation, based on the user model → algorithmic component of selecting the content
Content presentation – once you know what you want to show, how to effectively present the information in the user interface → interface

Page based approaches

Could be a description of the item, page, video, news item, etc
Adaptation mechanism – selecting most appropriate page
Closet approach to recommender system – a lot of predefined item, we select most appropriate, and we show
Process
- Selection – if content based, then we match the content, collaborative – we select based on what others have liked
- Presentation – how we order them, how you render them on the page (layout)

Example- KBS Hyperbook

Eg which uses page-based adaptation to illustrate the approach
About particular aspects of programming – depending on who the user is, appropriate content is shown → adaptive information resources
Adaptive navigational structure - All the pages possible for showing is shown in the navigation
- Colour – indicates the pages that are suitable for the user
Not shown features
- Adaptive trail generation – need to see this page then this
- Adaptive project selection – to select a particular problem for user to work on, and based on that user can study appropriate page → second way of showing a bundle of pages
- Adaptive goal selection – more related to user model – user can have several goals, and the system decide which goal is the most appropriate based on the goal, and it can give resources or paths to study THEN it can give adaptive project selection

Page-based approaches advantages and disadvantages

Advantages – versatile!!	Disadvantages – static!!
Widely used and is a mainstream approach – versatile approach Content ins pre-prepared → have a control of what the user will see, in businesses etc Easy to implement – you decide what to rec, and think how to order Can be used with many user models - if we consider content pages being items you see a lot of models used	Content authoring is expensive – need to have a dedicated team which will think what the best way to present the content is Efficiency of model being static – if new users come, they might not be catered for Content quality varies – depends on the authoring and the curation process. The curator might not know all contents → will cause deficiency of the problem

So the next approach will challenge the staticness

Fragment based approaches

We have predefined fragments, and we select parts and we bring the pre-defined parts together
Two strategies below!!

1. Optional fragments

Fragments could be anything – text, images, videos, etc – pieces of content that has been pre-prepared
Selection - Each fragment is associated with the applicability conditions (yellow little tag) !!!! the metadata that describes in which condition will be related so that we can pull the content
- Red – selected, based on its applicability conditions
Composition - Once selected, we need algorithm which constructs a page from several different fragments together
→ Have composed a page by pulling what’s available from the content pool

Optional fragments advantages and disadvantages

Advantages – flexible !!	Disadvantages – static!!
They are flexible. It gives element of dynamicity Fragments are pre-prepared, so we still have the control of what the user sees Fragment authoring is easier – rather than creating a fully pledged page !! Could be crowdsourced – since they are smaller pieces, it can be created by different authors → could speed up the process of content creation	Fragment quality may vary – as it is dependent on the author and how we've curated the content The pre-defined conditions for each fragment User may be missed – could be created when the user model isn't available yet Expensive/ time consuming – for every fragment, to create all the conditions Combined fragments may not align with one another, as it is from different authors

Altering fragments

Page is constructed as a set of constituents → offsets some of the deficiency
We don’t define preconditions for every fragment, but for group of fragments
- Content suitable for novice, intermediate or advanced
- Blue, yellow and red is associated with the level !!
Step – select, and put them together
1. Selection - We select the group (constituents) – and decide which group is appropriate for the user
  - After selecting the group, we go on fragments to find appropriate fragments
  - Called altering bc you can decide from the group, which to select
    - Can do it randomly, in order etc
2. Composition
  - Easier than optional, bc you can have layout of how you going to put the constituents
  - In e.g. put the red first then yellow !!

Altering fragments advantages and disadvantages

Advantages – flexible	Disadvantages
Flexible becausefragments can be pre-prepared (same as option) Easier than editing full content Can be crowdsourced Constituents allow control of coherence	Continues to be static!! becausecontent is pre-prepared Quality may vary Pre-defined constituents Authors need to think of what group fragment will go to Usually, the constituents are defined before the page composition or fragment definition → so definition of constituents narrow the condition for curators

→ even the most flexible approach, altering fragments, suffer from having to define the conditions when content is used

Part 2 – Dynamic approaches

Dynamic approaches– two steps

Dynamic content adaptation
1. Automatic selection of content
2. Automatic structuring of the content – decide which part is suitable and how to present
Dynamic content presentation
1. Automatically define relevance and focus – how to show what to focus and what’s in context
2. Automatic media adaptation – which devices used etc → much more complex, so will just illustrate the ideas

1. Dynamic content adaptation

Content automatically selected from:
- Knowledge base – need to have some relevance measure to determine which of the objects from knowledge base is suitable (e.g museum example- need to decide which painting and what info, and other painting to bring)
- Bayesian network – provides domain knowledge of what could be relevant to the user- gives a causal probabilistic relationship between the variables and can propagate from it, then decide which item is appropriate to show
- User preferences model – match the preference model to what we know about the world and identify which preference has the most importance → weighted preferences
Content automatically structured
- Task accomplished planners – bringing several pages together according to task and goals
- Argumentation models – you know what objects you want to show, and this model helps putting it together in a coherent manner to convince the user
- Conversation theories – brings interaction (dialogue) with the user to convince, or can present one narrative

Example – GEA (generator of evaluative arguments)

Using argumentative approach
First - Select what content is relevant
- Based on 3. Preferences – we identify which parameters are related
- E.g. type of house, location, number of bedrooms
- All the other parameters are suppressed
We need to convince the user that this is best for them – like a story line
- Story is building an argument to convince them
- E.g. convenient house location, and compromise for user preference, traffic is moderate (strength of the house to compensate for the deficiency)
- → adding more conditions related to house preferences & are generated automatically
Convince that this house is best for them
Level of detail will differ for users or for same user at different stages

2. Dynamic content presentation

Key is to decide what is the focus in the user, and what is the context
Focus – emphasize the content that has been found most relevant to the user
Context – allow access to less relevant content to preserve context
- Which user can go and stretch further
- Stretch text – you have item in focus are stretched out, in full detail – page cursor close or stretch !!!
  - Summary & purpose is stretched, but basic intro is closed
- Scaling fragments, dimming fragments, summary thumbnail

Scaling approach

How we present the focus and context – scaling
The fisheye – visualisation technique
- Is relevant but not the prime focus – is smaller and is blurred out.
- The prime focus – shown with full size and highlights so they can read and grasp quickly

Dynamic content presentation

How to adapt to different media factors
User specific factors – visual impairments (careful of size), dyslexia (more images), autism (no long list of bullet points)
Information specific feature
Constraint from media – if its speech, we need to take into account the user listening environment, video then bandwidth, textual, size of the screen and how much they can see
Contextual information – who else is around the user (other interactions)
Limitations of technical resources
→ need to bring those when we automatically generate and present it. Need to identify which are the most important ones for adaptations

2 widely used approaches of dynamic content presentation

Rule based approaches – using rules to define how to combine /choose different type of media in which condition
- E.g. dyslexia – structured diagrams and illustrative images > long text
- Tell us how to combine more than one media
Optimisation approaches – allow which items to put together, but how to best optimise the screen space and environmental constraints so that you can present most optimal media and media combinations

Example – RIA – Real Estate Agent (media adaptation – optimisation)

Focusing the presentation on the screen
How we select the houses and the characteristics of houses to show to the user
Then we need a mechanism on algorithm to tell us how to render the screen in a way media complements one another
- LHS – combination of several houses. On the screen and info is rendered so it isn’t overlapped
- Top side - complementary text → Optimisation – timing, which part of the dialogue and how it will correspond to images
- RHS - The $499000 is the key information now. Everything else is stripped out → optimisation is the level of detail & timing

Summary

The main adaption mechanism of the system – what to show and how to show
2 main family
- Static – predefined pages
- Semi dynamic – predefined fragments
- Dynamic – automatic content selection and presentation
Links to the recommendation algorithm, filtering and reordering of the contents