On IRobot--A Man And Computer Interaction System That Acts Upon A Quasi Natural Language

This paper can be divided into three parts mainly.The first part gives a brief introduction about the history of artificial intelligence research, and some recent approach to human-computer interaction system that's related with the natural language processing. The overview is simple because as a multi-discipline area, its study involves too many contents to be discussed in detail in a short and limited episode, which is not needed here. To review it emphatically, we later give a simple introduction about the progress in the study about language-based human-computer interaction system, which is what our system belongs to. Basing upon this, through comparing with the idea of Brooks about artificial intelligence study, our idea about it and how the whole paper arranged is made clear.The second part introduces some of our theoretical approaches to the problem. In this part, to draw our work out conveniently and plainly, we first introduce the idea and work of Albus about the "engineering of mind", after which, our idea and work are explained. In this process, we don't feel in a hurry to deny some classic and important theories or ideas about artificial intelligence research (e.g., the knowledge representationism), we would, on the other hand, consider about how to apply those theories or ideas in our recent study. For an example, basing upon a formal model of agent proposed earlier, an innovative idea about the unification and separation of the ego of agent is proposed by us, basing upon which an internal negotiation model of agent is proposed further. Since man's needs and desires show to be of multiple characteristics, this quite is hopeful for supplying us a more reasonable model of agent. For another example, through a detail analysis about the decision and behavior character of an intention system, we point out that pure logic system is unfit for the design of agent's decision or behavior, which should, on the other hand, be involved in some proper knowledge system. For this reason, we construct a knowledge system K_K on the basis of minimum modal logic system K to solve above problem. Mixed with the concept of dependent truth proposed by us, the system basing upon it can at least avoid the famous logic omniscience problem efficiently. Therefore, these work, as well as our some other work discussed later, can in some sense improve our architecture about complex and intelligent system.However, as what have been pointed out by many scholars, the development of practice is far behind the development of theoretical approach in artificial intelligence research, which makes the later of some special importance to recent research. The result of applying our theoretical approach to the problem is the development of iRobot, a human-computer interaction system that uses a quasi natural language as the media of communication, though it may be too simple to us man. From the third episode, we will go on to discuss about the meaning, function and characteristics of iRobot, which is the main content of the third part. The quasi language mentioned is defined after natural language English mainly. Even though it has just some simple syntax and vocabularies (they can be improved gradually), it has some similarity with English. Nevertheless, the system is endowed with some basic capability to learn basing upon some kinds of definition, which is an important way for human being to learn and grasp some useful knowledge. Comparing with the similar systems, it has also some particular advantages. It cares both the characteristic of natural language and its related function, e.g., it managed to grasp and apply such functional parts of natural language like "the","an" and "of etc, which makes great promise for the absorption of more functional characters of natural language in its definition later. Considering that it is in fact a real-time system, it is hopeful for applying in the robot control area (or other similar areas) where language-based communication, immediacy and some degree of intelligence are required.In addition, some programs mentioned in the paper are listed in the appendixes (more other programs are just omitted), which can be used as references. Confined to the time, there are many places to be improved, some of which are under studying now. And, since my ability is very limited, I hope to hear or learn more suggestions from experts of this area eagerly.