Intelligent Widgets

From Master Projects
Jump to: navigation, search


has title::Intelligent Widgets
status: finished
Master: project within::Software Engineering
Student name: student name::Vasileios Prountzos
Dates
Start start date:=2010/06/01
End end date:=2010/12/01
Supervision
Supervisor: Chris Van Aart
Second reader: has second reader::Rahul Premraj
Company: has company::The Widget Company TWC
Poster: has poster::Media:Media:Posternaam.pdf

Signature supervisor



..................................

Abstract

The Widget Company (TWC) is a full service company that is creating, developing and distributing widgets for multiple platforms (web, television, desktop and mobile).

Widgets are portable standalone chunks of code, which provide single – purpose services that can be installed and executed into third party sites by any user who has rights of authorship of the site. End users primarily use widgets to enhance their personal web experiences, or the web experiences of visitors to their personal sites. A rather common usage of widgets is their deployment in social networking websites to increase functionality and usability. In case of desktop and television platforms, end users may use widgets as interactive virtual tools to view on demand, capsulated information from predetermined data sources and access personalized content, based on the user’s preferences. Moreover, widgets provide a simple, comfortable and straightforward way for users to access certain services, which are application specific and predetermined in the functionality of a widget.

The primary objective of the project is to augment and supplement the current architecture and the current design of web widgets, in order to enable them to be intelligent and context aware. This means that the outcome of the project would be materialized into a logical component within the current architecture/design of widgets that would realize a series of algorithms that would gradually implement various levels of widget intelligence. In the first level of intelligence, an algorithm that would allow widgets to “understand” whether some user’s activity that takes place, is relevant to a widget’s services, thus this activity needs to be identified and further processed. In the second level of intelligence, classification algorithms based on supervised learning techniques would be investigated, which would automatically identify the type user’s activities that took place and “classify” them to the corresponding widgets, thus feeding the widgets with the appropriate information and data, for their operations. Following, the identification of activities, a series of actions on behalf of the widgets needs to take place. These actions are application specific and are based on the intentions of TWC regarding the type and quality of services that they vision for the widgets they develop. The knowledge engineering part, which consists of defining a rule base indicating the type of actions that need to be performed, after an activity has been identified as relevant for a widget, is not included in the scope of this project.

Problem statement: Assume that each widget is annotated or described by a set of keywords, or features that describes the service and the functionality that it provides. This feature set actually defines a class of concepts that are of interest to the widget.

Given a user’s social activity, we would like to identify in real time and with adequate precision, if possible, whether this activity is of interest to one or more widgets.

This approach is treating the problem of intelligent widgets that are able to infer improvements on their services and content, as a problem of identification of the relevant, from the widget’s perspective, activities, while taking into account a broad user’s context. Activities can be viewed as sentences or documents; thus the problem is a problem of document processing and falls into the referenced domain – Natural Language Processing. Various techniques from other domains (Artificial Intelligence, Machine Learning, Information Retrieval) can also be used, and are investigated, to process the online social activities of a user and to assist in the identification problem.