Marios Belk

EPL425: Internet Technologies

Technologies for Web-based Adaptive Interactive Systems: Personalization Categories, and Adaptation Mechanisms and Effects

Adaptation ComponentUser Modeling Component

videosimages

text

Decision Making & Adaptation Mechanisms

Adaptive User Interface

Usability User Experience

7 out of 40

High-level Adaptive and Interactive System Architecture

Agenda

Underlying principles of adaptation and personalization from a technical and design perspective

Technical perspective– personalization categories– adaptation technologies for adapting content and functionality based on the

characteristics of each user– how the Semantic Web and the Social Web contribute to AIS

Design perspective– adaptation effects that are communicated to the user interface for adaptation and

personalization systems

Reference selected state-of-the-art adaptation and personalization systems and frameworks

Personalization Categories

Link personalizationContent personalization

Personalized searchContext personalization

Authorized personalizationHumanized personalization

Link Personalization

Adaptation and personalization of the structure and presentation of hyperlinks in an interactive system

Achieved by selecting the links that are more relevant to the user (e.g., based on interests, preferences), changing the original navigation space by reducing or improving the relationships between nodes, and adapting the presentation of links

Popular in– E-Commerce– Educational Hypermedia Systems

Content Personalization

Adapting and personalizing the content of the user interface

Categories– Node structure personalization entails filtering the content that is

relevant to the users, illustrating sections and information in which the users may be interested

– Node content personalization is finer grained than structure personalization and involves adapting the information of the same node to various users

Personalized Search

The process of tailoring and personalizing the search results to an individual’s interests by taking into consideration information about the individual beyond the query provided

Implemented on the server side as part of a search engine’s methods or on the client side on the user’s computer (e.g., as a plugin on the Web browser).

Two general approaches to personalize the Web search results– By modifying the user’s query– By re-ranking search results

Personalized Search

Modeling users’ information for personalized Web search. This can be achieved through the following techniques:

– Personalized search based on content analysis in which the system compares and checks the content similarity between Web-pages and user models

– Personalized search based on hyperlink analysis in which the system computes the personalized importance of Web documents for each user

– Personalized search based on collaborative approaches in which the system presents similar search results to users with similar user models

Context Personalization

Adaptation of information that is accessed in different contexts of use

– User’s location– Interaction device– Physical environment or social context

Example– Text-recognition CAPTCHA mechanism may localize the text-based

challenge by presenting characters personalized to the users’ localized information

vs.

Authorized Personalization

Applied when an interactive system provides different access of information and action permission to users with different roles in the system.

Role-based access control: access rights in particular sections of a system are categorized under a role name. Most widely known approach

Team-based Access Control: collaborative team work and incorporates context information (i.e., the members of a team and the object instances) that is associated with collaborative tasks and accordingly applies this context information for access control

Humanized Personalization

Aims at creating personalized user interfaces based on intrinsic human factors

– Emotional factors (anxiety, stress)– Personality traits– Cognitive styles– Learning styles– Visual attention– Elementary cognitive processing abilities, etc.

Given the highly complex and multi-dimensional character of these factors, personalizing content and functionality of interactive systems based on such human factors is still at its infancy and not yet widely applied in commercial interactive systems

Adaptation Mechanisms

Adaptation Mechanisms

Implementation mechanisms to provide adaptation effects on the user interface based on the user model

Main adaptation mechanisms– Basic adaptation mechanisms– User Customization – Rule-based mechanisms– Content-based mechanisms– Collaborative-based mechanisms– Web mining– Demographic-based filtering

Basic Adaptation Mechanisms

Count how many times a node has been accessed

User Customization

The system provides an interface that allows users to construct a representation of their own interests

System is not considered adaptive but rather adaptable– But still this mechanism provides personalized content to the user

Rule-based Mechanisms

System has rules to adapt content and functionality based on the user model characteristics

Online Banking System

[USER].logged == False AND [USER].loginattempts.count > 2

Content-based Mechanisms

Extract keywords from documents the user has visited, bookmarked, saved, or explicitly provided

Assign weights on each keyword indicating the importance in the user model

Documents retrieved in response to search are also represented as a weighed vector of keywords

Compare the profile vector with retrieved documents’ vectors

Create User Model

Golf 0.3

Surfing 0.9

N top most frequently appeared keywords are included in the profile

Golf 0.2

Surfing 0.6

Football 0.4

Search

Adaptation Process

Display documents that are similar to the user model

Suggest the user links to a specified page by analyzing page content

Collaborative-based Mechanisms

Based on the assumption that if users X and Y rate n items similarly, or have similar behaviors (e.g., buying, watching), and hence will have similar interests

31 out of 40

Web Mining

Web mining includes data mining techniques with the aim to identify patterns from Web systems

Main categories– Web content mining which aims at the extraction and integration of

data and knowledge from Web-page content– Web-structure mining which aims at the analysis of node and

connection structure of a Web-site– Web usage mining which aims at extracting useful information from

server logs about the interaction activity of users, e.g., discover what users are looking for in a Web-page

Web Mining

Applies statistical and data mining techniques on server log data, resulting in a set of useful patterns that indicate users’ navigational behavior. Given the site map structure and usage logs, a Web usage miner provides results regarding usage patterns, user behavior, session and user clusters, clickstream information, etc.

Data mining methods employed– Association rule mining– Sequential pattern discovery– Clustering– Classification

Web Mining process– Pre-processing and data preparation, including data cleaning, filtering, and transaction

identification, resulting in a user transaction file– Data mining step in which usage patterns are discovered via specific usage mining

techniques

Demographic-based Filtering

Complements other adaptation mechanisms such as rule-based and collaborative filtering, aiming to refine the personalization result

Utilizes demographic information of users (e.g., age, gender, profession, etc.) to infer users’ interests and accordingly recommend particular objects

This method uses demographic information to identify the types of users that prefer a certain object and to identify one of the several pre-existing clusters to which a user belongs aiming to tailor recommendations based on information about others in this cluster

Semantic Web Technologies for AIS

Necessity to study and design the structure of meta-data (semantics) coming from the provider’s side

Aim: feed the adaptation mechanism with semantically enriched, machine-understandable information in order to adapt the hypermedia content based on the users’ models

Ontologies

Ontologies are widely used to organize and give meaning to information.

Ontologies formally define the types, properties and interrelationships of the main entities of an interactive system

SHOE: A rich and powerful representation ontology. A set of Simple HTML Ontology Extensions that allow Web authors to annotate their pages with semantics expressed in terms of ontologies

SHOE provides the ability to define ontologies, create new ontologies which extend existing ontologies, and classify entities under an “is a” classification scheme

The Semantic Web Initiative

Focuses on the creation of technologies and ontology languages and use of richer ontologies that can capture

a wider variety of relationship types that will facilitate machines to understand the semantics, or meaning, of information on the Web

Ontology representation languages – SHOE– Extensible Markup Language (XML)– Resource Description Framework (RDF)– DAML + OIL– Web Ontology Language (OWL)

High profile service providers such as Google utilize Semantic Web technologies by using RDFa and Microformats embedded in XHTML ( 2015 ), with the aim to support enhanced searching in Web-pages

– Usage example: Google states that the extra (structured) data will be used in order to get results for product reviews (e.g., CNET Reviews), products (e.g., Amazon product pages), and people (e.g., LinkedIn profiles)

Social Web for AIS

Researchers have investigated the effects of personality traits on human behavior in social networks or for implicitly eliciting personality traits based on social interaction data

Examples– Wald et al. (2012) utilized data mining and machine learning techniques to

predict users’ personality traits based on the Big Five personality model, using demographic and text-based attributes extracted from the users’ Facebook profiles

– Ortigosa et al. (2014) proposed an automated approach for predicting users’ personality traits based on Facebook usage Facebook application which has been used in a large scale study ( N = 20,000) to

collect information about the personality traits of users and their interactions within Facebook

Classifiers have been trained aiming to identify interaction patterns (e.g., number of friends, number of wall posts) that relate to (and eventually predict) users’ personality traits

Adaptation Effects

Interactive System Architecture - What to Adapt?

Interactive Systems

Information Architecture Functionality

Content Content Presentation

Content Navigation

What to Adapt?

An important adaptation issue is which visible features of the system can be adapted by a particular technique

Two main classes of adaptation technologies– content-level adaptation, called adaptive content presentation– link-level adaptation, called adaptive navigation support

Adaptive presentation relates to the adaptation of hypermedia elements inside nodes

Adaptive content navigation support relates to the adaptation of links inside nodes, indexes and maps

Adaptation Effects in UIs

Content-level adaptationAdapt the hypermedia elements (or content fragments) of a node

Link-level adaptationAdapt the presentation of hyperlinks within a node in order to support user navigation in the hyperspace

Content-level Adaptation

<SELECT statement> ::= [WITH <common_table_expression> [,...n]] <query_expression> [ ORDER BY { order_by_expression | column_position [ ASC | DESC ] } [ ,...n ] ] [ COMPUTE { { AVG | COUNT | MAX | MIN | SUM } (expression )} [ ,...n ] [ BY expression [ ,...n ] ] ] [ <FOR Clause>] [ OPTION ( <query_hint> [ ,...n ] ) ] <query_expression> ::= { <query_specification> | ( <query_expression> ) } [ { UNION [ ALL ] | EXCEPT | INTERSECT } <query_specification> | ( <query_expression> ) [...n ] ] <query_specification> ::= SELECT [ ALL | DISTINCT ] [TOP ( expression ) [PERCENT] [ WITH TIES ] ] < select_list > [ INTO new_table ] [ FROM { <table_source> } [ ,...n ] ] [ WHERE <search_condition> ] [ <GROUP BY> ] [ HAVING < search_condition > ]

Link-level Adaptation

User Features

Knowledge, Background, Individual Traits, Location

Interests Knowledge, Background, Individual Traits

Knowledge, Background, Goals, Individual Traits

Interests, Location

Domain Educational Hypermedia Systems News systems, Online Commercial Shops

Educational Hypermedia Systems

Educational Hypermedia Systems, Web‐based systems

Web Recommender systems

Aim Guide Reduce navigation time and steps

Reduce cognitive load Support navigation Reduce navigation time and steps