Brought to you by EarthWeb
IT Library Logo

Click Here!
Click Here!

Search the site:
 
EXPERT SEARCH -----
Programming Languages
Databases
Security
Web Services
Network Services
Middleware
Components
Operating Systems
User Interfaces
Groupware & Collaboration
Content Management
Productivity Applications
Hardware
Fun & Games

EarthWeb Direct EarthWeb Direct Fatbrain Auctions Support Source Answers

EarthWeb sites
Crossnodes
Datamation
Developer.com
DICE
EarthWeb.com
EarthWeb Direct
ERP Hub
Gamelan
GoCertify.com
HTMLGoodies
Intranet Journal
IT Knowledge
IT Library
JavaGoodies
JARS
JavaScripts.com
open source IT
RoadCoders
Y2K Info

Previous Table of Contents Next


4. RESEARCH ISSUES

The representation problem is an endless issue because no final solution is under the technological horizon for creating a complete identity of the world and of computer representation. The most ambitious hard advocates of artificial intelligence speak only about the substitution of the human mind, though this is also a dubious expectation. The human mind itself is not the representation body of all possible knowledge, the complexity of the world is much beyond the complexity of the mind.

In the course of representation process the first research task is to find better means for knowledge acquisition. The problem is more rooted in cognitive psychology but the lessons of better human cognition and communication about knowledge are issues to be implemented by the computer representation, especially in the forms of input representation. Icon and other graph representation forms of nowadays common software product were results of earlier computer representation research. This development is only at the beginning; the question how to design flexible, interactive knowledge input interfaces for experts and for non-expert users, for different age groups and cultural environments, for everyday, nearly continuous use or for accidental man-machine communications is still completely open.

The other input-related representation problem is the natural language interface. Natural language contrasted to computer languages contains lexical and semantic ambiguities, sollipsis, i.e., omissions which suppose a background knowledge of the partner and her/his active participation in the discourse, metaphors anticipating the same cultural background of the other side, sentences which are not correct syntactically, jumps in the logical sequence of the subject, uncertain, not well organized references within the text. Verbal communication is understandable within a certain environment which defines the pragmatics, i.e., the representation of the communicators and of the communication, considering the situational relativity of the communication, meaning. A reliable and complete understanding of natural spoken language is not either attainable between two persons. The machine solution is farther, though limited results are present, understanding limited subject areas with limited vocabulary, terms, and they can be very helpful in practical applications. The existence of many professional and familiar vernaculars strengthens the need for some common metalanguages, metarepresentations, such as the explanation texts of the best dictionaries. A steady progress is ongoing in all these directions.

How central the linguistic issue is, can be felt through the recent efforts of one of the founders and most influential researchers of artificial intelligence, McCarthy. He represents the whole problem in a framework of speech acts, i.e., knowledge-based representation of communication among agents. Somehow similar idea is behind Schank's research on goal-based scenarios.

The next research issue is the extension and combination of different representation methods. Some higher order logical means expressing embedded knowledge can be more powerful than present first-order logic, straightforward representations. Reasoning based on higher order logic is clumsy, sometimes unsolved. Further combinations of uncertainty methods and logic, their automatic or semiautomatic matching with the most analogous methodological models are future problems.

One of the most promising and fast progressing representation trends is the application of connectivity ideas. This is done in many directions, first of all in hardware. The latest achievements in cellular neural nets provide a several-thousand time faster access to a whole image than any chip before. The image, the pattern can be used as a direct representation of some visual information but also as conceptual structures, entities, similar to the philosophy that was represented in this chapter through the relations of metapatterns and concepts. Physical and programmed links between logic based representations and connectivity-related patterns are emerging. These will help in future development of space and time representation. The human mind's space and time representation history followed somehow in a similar way. A general and automatically given space and time representation can be a frame for most future machine representations; the rapid advancement of multimedia proves this relevant cognitive feature.

5. FUTURE TRENDS, CONCLUSIONS

The general trend of artificial intelligence and related research is a fast unification with all other trends of computer science and practice. Many of the esoteric considerations and imaginative methods of the past are now natural ingredients of the most common products if they were feasible indeed. According to a witty remark: the real successes of science are those which are not noticed anymore but work. This is the relevant trend for a kernel of all software products like representation. The paragraph on input representations outlined the main directions in the interface field. These interface representations are now the hottest problems of practical user-friendliness. The interior representation trend has two parallel and only seemingly diverting directions. On the first, extremely ambitious projects, like the early General Problem Solver or the long dragging Cyc effort try to create very general problem solving, and behind that, try to create very general representation vehicles. On the other second highway, thousands of pragmatic information-, expert systems are born every year for specific use. As it was emphasized, a general, human-like intelligence, and a brain imitating representation scheme is non-existing on the horizon but completely isolated small systems cannot survive too long, and cannot be economical. The necessary integration follows the history of general computing: a few, open-ended operation systems created a frame for free cooperation of an abundance of products. This should be the future of the most successful, open expert system shells, application toolkits. The emerging tools create an environment where the usual task is not related to programming, computational, representational system design but is a steady exercise for domain experts, system analysts, cognitive psychology-oriented knowledge engineers. They all must have a certain general knowledge, but not more, about the interior nature of representation. All further computational work should be executed by the professional toolkit product.


Previous Table of Contents Next

footer nav
Use of this site is subject certain Terms & Conditions.
Copyright (c) 1996-1999 EarthWeb, Inc.. All rights reserved. Reproduction in whole or in part in any form or medium without express written permission of EarthWeb is prohibited. Please read our privacy policy for details.