I intend to write an expert system program to operate in the domain of technical support problem solving. In particular, this system will work in the sub-domain of certain problems that arise in the use of a particular software product. The product I have in mind to create this system for is one of the interconnectivity products produced by Sybase at their new center in the University of Colorado Research Park. Besides being a major relational database vendor, Sybase now has an interconnectivity division that is based here in Boulder. These products have the purpose of providing connectivity between various desk top database applications and the various back end databases that normally house a corporation's main data store. Part of this task involves translating the various flavors of SQL that each desk top vendor's database understands to the SQL version that is supported on the back end system. The other major task that is addressed is in the area of translating the various network protocols by repackaging the query data from one protocol, say IPX, to another, perhaps TCP/IP.
The main goal of my project is to design a series of experiments that would allow us to investigate the differences in strategies (algorithms) developed by humans after they are introduced to various visual presentations of the same task. The presentations will emphasize or de-emphasize the topological and geometrical (symmetric) aspects of the task. My choice of topological characteristics comes from arguments of proponents of the non-algorithmic models of human intelligence mentioned above. In choosing the symmetric characteristics, I follow Margaret Boden [1][2] who often emphasizes the role of geometrical images (in particular symmetry) in modeling human creativity. My ideas were also inspired by an experiment discussed in class concerning two different presentations of the game of Tic-Tack-Toe. However, in my project I intend to consider more complicated tasks (e.g. Sim game, etc.) and to explore a much richer variety of possible presentations.
What is the best knowledge acquisition technique wherein your likelihood of a successful application implementation is the greatest? This is the problem I will be addressing for my project. I am going to approach this problem from three different perspectives (academic direction, implementation examples, and personal experience with a specific problem domain).
Given many different attributes about people in our United States population, I would like to use a genetic algorithm (GA)-based pattern recognition algorithm to try to predict what combination of these attributes determines whether a person has a salary of greater than $50,000 a year.
Our project is to implement a system that uses qualitative sign algebra to solve systems of equations. We will use the sign algebra as described in the Brian C. Williams paper "MINIMA A Symbolic Approach to Qualitative Algebraic Reasoning", Proceedings of AAAI-88, p 264-269. The implementation will be done in MacScheme.
Several years ago I designed a state machine in a PLD that handled all the handshaking on the peripheral side of a Centronics bus. Occasionally when making state transitions in which two or more state bits changed, the state machine became stuck in an intermediate, undefined state. I spent many hours adjusting state bit assignments, until I finally resolved the issue.
This project will focus on natural language processing. The goal of this project is to study how the computer can be programmed to understand a story. The theory is that there is a script to every story in this world. Each script consists of one or more events. Each event has many components such as what can happen, who is involved, where can it take place, etc. When we are told a story, even we are only told only the partial story, we can fill in the missing events and have a complete picture of the whole story. Therefore, a program can be written so that it has some pre-defined scripts with events. When this program is given a story, if its events match some or all the events in a script, then the story is understood even if the input story is missing some events.
To my knowledge, no computer has ever been programmed to play the game of scrabble. I plan to change this by producing an interactive program which takes, as input, the current state of the board and outputs the state plus one move.
In our project we are studying the usefulness of collaborative agents as a means of implementing evolving or adaptive software. We use the term adaptive software to describe software that relies primarily on feedback from within the system, with little or no user feedback. Other research has explored collaborative agents in term of their ability to share information. We will take this idea further by creating an environment where the usefulness of an agent s knowledge and skills is a key factor in determining whether or not those traits get perpetuated. Likewise, knowledge and skills that prove to be less useful in a given environment will become extinct through what one might call natural selection. We use the term knowledge to describe meaningful information that is used by an agent to understand its environment. The term skill is used here to describe a technique for achieving a goal. The goal in this environment is survival.
Valentino Braitenberg in his book RVehicles, Experiments in Synthetic PsychologyS [1] describes a series of thought experiments with the intention of conveying his philosophy on the problem of the mind. This journey is done not through a Rworld of real brains but through a toy worldS that he creates along the way, by conceptually TbuildingU machines with very simple internal structures that can be thought of ones that exhibit behavior. Braitenberg concludes the introduction to his book by inviting us to Rget used to a way of thinking in which the hardware of the realization of an idea is much less important than the idea itselfS. Despite this remark, my project will attempt to realize some of these ideas in order to experiment with them in a hands-on way rather than the thought-experiment way.
My project is to design and implement a Machine Translator from a Controlled English Language queries to formal database SQL language in a well-defined domain. The objective of this project is to develop a Controlled English Language Interface (CELI) to a database system (for example, Students database). This interface provides a common medium of communication between a person unfamiliar with the computer languages and internal storage mechanisms and will be able to access the information stored in the database. (ref: NLI).
We will investigate the use of AI agents in the domain of resource scheduling in a homogeneous distributed operating system environment. We are interested in understanding the difficulties in applying intelligent agents to resource scheduling in general. Defining appropriate functionality for Resource Scheduling Agents (RSA's) will be our primary goal. It is our intent to limit our approach and focus on designing RSA's for allocating processes to processors efficiently and transparently to users within a networked workstation environment.
Writing is a difficult task. Successful writers have established their vocabulary and style through years of practice and experience. Some of the basic tools of their trade are the dictionary, thesaurus, and style manuals. Within the past decade, additional tools have been developed in the form of computer applications to check spelling and even check grammar. They find misspelled words or incorrect grammars and provide alternative spellings or grammars so the writer may pick the appropriate one so that correctness is ensured. What if instead of just checking for correctness of a sentence, such an application provided some alternative colorful sentences that convey the same theme? Such a sentence thesaurus could be a good tool to beginning writers (or at least inexperienced elementary school students) by providing alternative sentences, possibly lodging loose some creativity out of their cluttered mind.
In my project, I plan to use Genetic Algorithms to accomplish some specific tasks in music analysis and production. Using AgentSheets (and Visual AgenTalk) as a framework, I will design and implement algorithms that operate on musical input and produce musical output.
Unlike English, Chinese can not be directly inputted via keyboard because it is not an alphabetic writing system. A conversion program is needed to convert into the corresponding Chinese characters the pronunciation which can be represented by English alphabet phonetically.
The purpose of this project is to develope a system which provides a pen based interface to existing mouse and keyboad interface applications. The system translates userB%fs handwriting symbols into actual inputs to the applications. The translations from symbols to inputs are modifiable by the users. We will study how effectively pen and tablet based interfaces work in comparison with the environmentB%fs traditional inputdevices. We assume that the pen and tablet will be used in conjunction with mouse and keyboard input devices. Also we will study how users modify the symbol translation table to fit their purposes and preferences.
I wish to explore the topic of Programming by Analogy in order to provide an intuitive mechanism that makes it easy for end-user programmers to reuse code. I plan to learn much about ways to present analogies to people in ways that are fairly intuitive. Analogy and metaphor seem to be deeply rooted in our ways of knowing the world (Lakoff, 1987; Hofstadter, 1991). I think it will be interesting to explore these ways of understanding the world in the context of programming. If I can come up with a mechanism that captures even a bit of the powerful analogy-making capabilities that people seem to exhibit, it may become much easier to create agents (within a simulation environment) that share behavior.