User interface adaptation based on user feedback and machine learning

Editor's Notes

• #2: Welcome to my presentation about user interface adaptation based on users feedbacks and machine learning, my name is nesrine mezhoudi , I’m a PHD student and a research assistant in the catholic university of Louvain , I’m investigating the potential of machine learning for adaptive user interface and specially model based User Interfaces.
• #3: Nowadays, Context ware user interface, wish denote the UI adapted to their context of use , are holding a big interest. given that the context of use is defined by the triplet<user, platform, environment> ; user, denoting the end user, platform representing the device holding the user interface and the environment in wish the interaction is taking place. currently , current systems migrate the adaptation of user interface into more user centered aspect focusing on the end user preferences and aiming at increasing their usability as well as their user satisfaction and user experiences.
• #4: Nowadays, Context ware user interface, wish denote the UI adapted to their context of use , are holding a big interest. given that the context of use is defined by the triplet<user, platform, environment> ; user, denoting the end user, platform representing the device holding the user interface and the environment in wish the interaction is taking place. currently , current systems migrate the adaptation of user interface into more user centered aspect focusing on the end user preferences and aiming at increasing their usability as well as their user satisfaction and user experiences.
• #5: Nowadays, Context ware user interface, wish denote the UI adapted to their context of use , are holding a big interest. given that the context of use is defined by the triplet<user, platform, environment> ; user, denoting the end user, platform representing the device holding the user interface and the environment in wish the interaction is taking place. currently , current systems migrate the adaptation of user interface into more user centered aspect focusing on the end user preferences and aiming at increasing their usability as well as their user satisfaction and user experiences.
• #6: I’m going to start by giving an overview about this presentation, it is structured as follows , First I’m going to introduce the concepts of adaptive user interfaces and contextualization, beside the context of use and adaption which are considered the key points for a user centered design . Then I will put in evidence the weakness of existing works considered as the main motivation for the current investigation of machine learning techniques for user interface adaptation. Later we introduce the basic concepts and the background of the works before outlining the approach and its forseen application Finally I will conclude and give the outlooks of this investigation.
• #7: Commonly Adaptive user interface are based on a set of predefined and even static adaptation rules defined regarding standards, ergonomic guidelines, esthetic metrics… Accordingly it is hard to readapt the UI neither to update the existing adaptation rules wish supposed to evolve across time. and even it it is possible in what costs ? this lack of dynamic adaptability among systems has several consequences: user dissatisfaction because these systems do not fit the users needs, limitations, and preferences, user demotivation to use the system and finally user withdrawal Thereby the dynamicity and the interactivity od adaptation began a crucial requirement to keep the user and ensure their satisfaction.
• #8: Commonly Adaptive user interface are based on a set of predefined and even static adaptation rules defined regarding standards, ergonomic guidelines, esthetic metrics… Accordingly it is hard to readapt the UI neither to update the existing adaptation rules wish supposed to evolve across time. and even it it is possible in what costs ? this lack of dynamic adaptability among systems has several consequences: user dissatisfaction because these systems do not fit the users needs, limitations, and preferences, user demotivation to use the system and finally user withdrawal Thereby the dynamicity and the interactivity od adaptation began a crucial requirement to keep the user and ensure their satisfaction.
• #9: Thus , systems need to converge to advisory software considering the user intervention as a highly valuable factor for adaptation. Accordingly , Interaction shouldn’t be overlooked (neglected ) as a factor for adaptation as well as the context of use in order to get a more appropriate user adaptation. Wish seems promising to enhance the users preferences influence in the user interface definition and adaptation moreover for the update of the adaptation rules set and their management. To meet such need Machine learning techniques seems potential and promising .
• #11: During this cycle user behaviors and intervention provide a rich source improving the quality of adaptation , usually user feedbacks are classified into implicit and explicit feedbacks The implicit one consider that every interaction as an interest indicator but those feedbacks class still enable to to illustrate a dislike attitude of users. The explicit feedback consist in acquiring knowledge by asking user in more expressive way . Here we present the principal topologies to express feedback which consider mainly too axes implicit/explicit and with or without rating aims .
• #12: With regard to above mentioned cycle , we propose a theoretical architecture unifying machine learning based adaptive systems . it convey the main modules unifying the conceptual description of various possible instantiation. A ml based Adaptive user interface is mainly structured around an instantiated context of use( user, on a platform in an environment). Beside it focuses on the exploration of the interaction flows with users since that provide advantageous data. The exploration of interaction consist in two step; the firs is data perception: when the system track the users feedbacks The the feedbacks management via machine learning techniques and basically three module Reinforcement, Evaluation and update module Reinforcement consists on a process of consolidation of decision basing on user behavior; a clever exploration of behavioral pattern is required to ensure an accurate reinforcement. Evaluation is more based on calculating the gap between user attempts and system recommendations in order to optimize the quality of approvals. Both evaluation and reinforcement exhibit the learning purposes. They trigger the updating module by injecting new rules or modifying existing ones, in term of effects and priorities, according to user preferences deduced during the interaction .
• #13: The adaptation management layer is based on extraction and classification of rules and patterns Whish are the most popular ml techniques serving as a ‘Rule learners’ , based on supervised learning approaches, which is basically a rule management systems tools treating the users feedbacks and the set of existing adaptation rules to generate Learns new decision rules , which can be added to n the adaptation rule repository after being evaluated according to a promoting demoting techniques which give to the rule manager a new dimension of self improving learning capabilities.
• #14: The adaptation management layer is based on extraction and classification of rules and patterns Whish are the most popular ml techniques serving as a ‘Rule learners’ , based on supervised learning approaches, which is basically a rule management systems tools treating the users feedbacks and the set of existing adaptation rules to generate Learns new decision rules , which can be added to n the adaptation rule repository after being evaluated according to a promoting demoting techniques which give to the rule manager a new dimension of self improving learning capabilities.
• #15: The adaptation management layer is based on extraction and classification of rules and patterns Whish are the most popular ml techniques serving as a ‘Rule learners’ , based on supervised learning approaches, which is basically a rule management systems tools treating the users feedbacks and the set of existing adaptation rules to generate Learns new decision rules , which can be added to n the adaptation rule repository after being evaluated according to a promoting demoting techniques which give to the rule manager a new dimension of self improving learning capabilities.
• #16: The second aims of this research is to avail adaptation based on learning for model based user interfaces , and basically in different levels of UI generation according to the CAMELEON reference framework ]. The focus of this point is to capitalize on learning based adaptation to improve the validity of the Abstract User Interface. Since , EVEN THAT Several algorithms were defined for the AUI definition however a lack of validity and consistency control still arise . Moreover, an intended reward of ML technique’s potential to be envisaged for the concrete user interfaces . Although some ML techniques were already explored , ML is consistently promising to emphasize new scenarios for the adaptation in the concrete UI level, for instance for the widget selection ML techniques seems great to manage all adaptation rules and guideline.
• #17: The second aims of this research is to avail adaptation based on learning for model based user interfaces , and basically in different levels of UI generation according to the CAMELEON reference framework ]. The focus of this point is to capitalize on learning based adaptation to improve the validity of the Abstract User Interface. Since , EVEN THAT Several algorithms were defined for the AUI definition however a lack of validity and consistency control still arise . Moreover, an intended reward of ML technique’s potential to be envisaged for the concrete user interfaces . Although some ML techniques were already explored , ML is consistently promising to emphasize new scenarios for the adaptation in the concrete UI level, for instance for the widget selection ML techniques seems great to manage all adaptation rules and guideline.