Personalized Intelligent User Interfaces

Competence Center: Information Retrieval and Machine Learning  
Contact: Prof. Dr. Sahin Albayrak 
Partners: Deutsche TelekomT-Mobiledpa


The acceptance and success of future communication services depends to a large extent on their usability. Users would like to use equipment and services which they can quickly and easily adapt to their own interests. The flood of information available on the Internet continues to increase, so there is a great need for a personalized filter for this information. With personalized, intelligent user interfaces, this need can be met. On behalf of Deutsche Telekom Laboratories, in the Personalized Intelligent User Interfaces (PIUI) project, technologies and options for the personalization of user interfaces are being evaluated. Criteria for differing types of interfaces and for promising areas of personalization are being defined. The needs and habits of potential users are also being analyzed.

Personalization Framework

For the project, a framework was specified for the explicit and implicit personalization and as then implemented as a prototype. Applications created like this adapt to each situation that a user gets into, as well as to their general preferences and abilities. This framework makes the personalization possible in many applications and via web services it can be integrated into a variety of applications on the user’s device. With the existing, partially specified solutions, this was previously not possible. Personalization services can thus make an important contribution to improving usability and thus reinforcing customer retention.

Feature Extraction Framework

Another framework, implemented as a prototype, is the Feature Extraction Framework (FEF). The FEF is a framework for handling text documents and their textual features. Textual features can be anything from a word appearing on a page, a key phrase that has been assigned or extracted, an identifier of a WordNet set of synonym numbers, the document’s author, a reference to another document, the document length and so on. The FEF handles all logics of extracting and assigning features to the documents. The FEF always considers documents in a special context by handling documents in sets called Document Collections. Saving these features along with each document will facilitate the process of clustering documents.

Concrete Scenarios

On this basis, a set of applications has been implemented as demonstrators to show the possibilities and benefits of an intelligent personalization system for user interfaces. E.g., with the help of a configuration widget, via a web interface, users could thus easily and intuitively set up their mobile devices and add new services using drag & drop. After configuring, the user interface on the mobile phone is automatically modified. On the basis of the user’s behavior, the system can also suggest new widgets (implicit personalization). Another example shows how intelligent user interfaces can be adapted to an IPTV environment. When selecting these, particular attention was paid to the commercial potential and to the approaches for possible integration into Deutsche Telekom services.