Mladen Stanojevic

Knowledge representation with SOUL

Good knowledge organization is crucial in two cases-in the representation of semantically related information and in the representation of huge quantities of data. The answer to the problem of semantic knowledge representation is given by various knowledge representation techniques used in Artificial Intelligence, while for the representation of large quantities of data, relational databases represent the optimal choice. If we analyze the characteristics important from the knowledge organization point of view, we could say that structure representation provides means for knowledge representation, unique object representation ensures representational efficiency, while representation of bottom-up hierarchies (e.g. indexing scheme in relational databases) brings search efficiency. Furthermore, unique object representation and bottom-up (contextual) hierarchies provide support for semantically organized knowledge. In this paper we describe a knowledge representation technique, Hierarchical Semantic Form (HSF), which, together with the Space Of Universal Links (SOUL) algorithm, aims at supporting both of the above mentioned characteristics of well organized knowledge. HSF represents a hybrid solution that enables both structure and context representations, combining the characteristics of classicist approach (symbolic and semantic processing) with the characteristics of connectionist approach (parallelism, learning).

A blackboard framework on top of prolog

Abstract BEST (Blackboard Expert Systems Toolkit) aimed to provide the ability to produce large-scale evolvable, heterogeneous intelligent systems. Apart from providing all necessary blackboard infrastructure and control capabilities, BEST allows its user to integrate different programming paradigms within a single blackboard application. It does not matter if one knowledge source is a forward- or backward-chaining rule-based system, another uses model-based reasoning paradigm, and still another is a conventional, procedural program. Moreover, because representational flexibility is similarly important in blackboard systems, and the blackboard model does not place any prior restrictions on what type of information can be placed on the blackboard, the authors offer multiparadigm knowledge representation language, built on top of Prolog, therefore avoiding restricting users to a single way of expressing either knowledge or data. BESTs problem-solving facilities include assumption-based truth maintenance, temporal, hypothetical, and approximate reasoning. The generality and flexibility of the BEST environment present the first-time blackboard application developer with numerous implementation choices, providing a wide range of options and capabilities,what significantly increases the productivity of BEST programmers, and improves the performance of the application they produce. The ability to explain its reasoning and to defend its decision also distinguished BEST from most other blackboard systems.

Investment Decision Making

Publisher Summary For large, diversified, multiproduct companies, investment decision making is an extremely complicated and multidimensional process. One of the main shortcomings in investment decision making is reliance on a single method and to rectify this, a multi-paradigm approach is recommended. An investment decision making expert system must aid the project analyst and investment decision maker to determine whether a project is acceptable; if it is, whether it is the best alternative; and to calculate the extent of the decision sensitivity to certain critical assumptions. Numerous measures of the financial profitability of each alternative can be calculated using standard financial tables based on an integrated documentation system. A spreadsheet seems to be the most relevant platform for this analysis. A full analysis should be done for major project alternatives, and the summary table should be transferred to the expert system that would perform a heuristic classification of the alternatives, and a multi-criteria analysis of the most promising projects, with the different weights assigned to various key objectives. These weights can be varied or refined according to the decision makers wishes and customer profile. Sensitivity analysis and other risk-bearing methods could be employed so that all impacts of imprecise forecasts on future benefits and costs can be seen in a proper perspective. For the purpose of the more extensive sensitivity analysis with respect to various types of benefits and costs, various degrees of change in key variables, and alternative timings, including the delay of the project completion and the delays in reaching full production, additional data could be included in an alternative projects summary table.

Semantic Classifier for Affective Computing

One of the most important fields of affective computing is related to the hard problem of emotion recognition. At present, there are several approaches to the problem of automatic emotion recognition based on different methods, like Bayesian classifiers, support vector machines, linear discriminant analysis, neural networks or k-nearest neighbors, which classify emotions using several features obtained from facial expressions, body gestures, speech or different physiological signals. In this paper, we propose a semantic classifier as a new, simple and efficient approach to the problem of automatic emotion recognition. The implementation of the semantic classifier is based on the basic, natural principles used to decrease the complexity of problems found in n-dimensional spaces: discretization, structure identification and semantic optimization. The proposed classifier exhibits some self-organizing features and supports learning by repetition, generalization and specialization. It will be used to implement a distributed and robust system for emotion recognition.

Semantic web services with SOUL

Semantic Web Services should make it easier for a user to find the needed information on Web by using natural language queries, instead of simple keywords like in search engines. It has been widely recognized that the main problem in the implementation of this idea is the problem of semantic representation, the same problem that AI researchers were trying to solve for a long time. Various ontology and schema languages are used in Semantic Web to represent the semantics of Web pages, but they require an extensive effort to translate the existing Web pages. We propose a new knowledge representation technique, so called Hierarchical Semantic Form, together with a supporting SOUL algorithm, which should provide a rudimentary understanding of existing, non-annotated Web pages, thus eliminating the need for their laborious translation. As an example we have implemented a prototype Semantic Web Service that gives information about flights stored in an ordinary Web page.

Blackboard metaphor in tactical decision making

Abstract The objective of this paper is to give an overview of some work we have done on tactical decision making within a blackboard framework. A dedicated simulation facility to provide a military commander with the capability for two-sided interactive air war gaming has already been built under support from the Yugoslav Army. To complete the training environment an automated gaming scenario generator was lacking. Therefore, we try to provide an intelligent decision aid to shorten the time needed for game preparation and to make the tool suitable for future commander training in tactical decision making.

A Decision Support System for Building Evacuation Based on the EMILI SITE Environment

This article introduces a decision support system for determining a safe and short (in both time and distance sense) evacuation path and for increasing situation awareness in the case of a hazard in buildings of large critical infrastructures (CIs). CIs are infrastructures that are complex, difficult to monitor and with an integrated SCADA system, such as airports or metro stations. The FP-7 project Emergency Management In Large Infrastructures (EMILI) aims at a new generation of data management and control systems in CIs [1] by building a further data management layer on already existing systems. Part of the EMILI project is the implementation of a simulation and training environment (SITE). The introduced decision support system is meant to be integrated into SITE, using results from event processing and an optimization algorithm to determine evacuation paths and increase the awareness of the safety situation. The decision support system was developed for an airport use case.

Application of Complex Event Processing Paradigm in Situation Awareness and Management

This paper describes a novel approach to situation awareness and management in critical infrastructures using an emerging Complex Event Processing paradigm in combination with ECA (Event Condition Action) rules. Every modern infrastructural facility, especially critical ones, would want to catch or predict exceptions and threats at the earliest possible moment. The crucial prerequisite for this is a holistic and accurate situation assessment. To be able to assess the situation, evaluate the risk and provide decision support to the emergency managers, we need to define event chains that identify what is normal, compliant and expected, and what is exceptional and/or dangerous. These problems introduce a novel genre of applications - event-driven applications that make automated decisions based on a complex event or pattern of events, their detection, correlation and aggregation. In cases where an automated reaction is not possible, a recommendation is given to a human operator who remains in the control loop.

Self-Adaptive Prototype for Seat Adaption

Self-adaptive prototype for seat adaptation aims at enhancing the physical comfort of a driver by taking into account not only the state of the environment (state of the road, car settings), but also the driver’s emotional, cognitive and physical state. To implement this prototype we used a REFLECTive middleware, which provides a programming framework for the development of pervasive-adaptive applications. The REFLECTive middleware supports self-adaptive behavior and is generally composed of three tiers: Tangible tier contains services that read sensors data and send commands to actuators, REFLECTive tier is responsible for analyzing the data collected from sensors and for defining the actions that will be performed by actuators, Application tier facilitates high-level decision making. The seat adaptation prototype uses the information about Center of Pressure (COP) speed and number of bumps to determine the driver’s physical state, and then it combines this information with the driver’s cognitive and emotional state to figure out if the driver feels uncomfortable, and to change the state of seat cushions in an attempt to make driver feel more comfortable. The components of the seat adaptation prototype in the REFLECTive and Application tier are implemented using reaction rules.

NIMFA - Natural language Implicit Meaning Formalization and Abstraction

There are many general purpose Knowledge Representation, Natural Language Processing and Information Extraction techniques that were successfully applied in many applications. However, their more broad use is still limited by the relatively high costs of their application. It seems that these limitations are partly caused by some essential characteristics and some weaknesses of these techniques. In this paper we propose a radically new knowledge representation and interpretation technique, NIMFA, specialized for knowledge expressed in natural languages. To test the basic ideas underlying NIMFA we have implemented a prototype Information Center that provides answers to natural language queries using Web services.