Controlled language interfaces to intelligent rooms

OData support
Supervisor:
Dr. Mészáros Tamás Csaba
Department of Measurement and Information Systems

Natural language communication is one of the many research topics on the field of artificial intelligence, understandably so, since the ability to communicate with words was a great step in human evolution, too. Natural language communication with machines is not only a more comfortable way of passing information, but it is also of great importance in some practical cases. Providing natural language based means of control for electronic appliances can help the aid of the elderly and the sick.

The computational modeling of natural language communication is a very complex problem which can be approached from different viewpoints. It’s possible to provide an interface with a limited number of interpretable natural language sentences, but that is not a dynamic base for communication. Studies on the field rather focus on developing intelligent systems which are able to communicate with the use of thought processes similar to that of humans. These systems are capable of understanding sentences and producing appropriate answers, but they require deep and continuous learning. While this seems to be the appropriate approach regarding intelligent agents, it’s not yet a solution which could be easily implemented in any kind of small applications.

In my thesis I explore the possibilities of providing efficient natural language communication between men and machines by restricting the communication to a smaller part of the language. I present a grammar system which is capable of dynamically handling natural language inputs based on limited grammar definitions. The system includes a general purpose parser which can adapt to the given set of interpretable structures. I analyze these topics through the development of an application which provides natural language interface for the controlling of an intelligent room. To provide a practically useful approach to the topic, I review the possibilities of creating a voice input based interface.

My paper focuses on implementing an Android application that provides a voice input based natural language interface for intelligent tools that use the X-10 industry standard. I examine the available speech recognition service, the capabilities of the X-10 based home automation and present the development of an easily customizable grammar system.

Downloads

Please sign in to download the files of this thesis.