Search

Due to growing interest in using natural language interfaces, it is appropriate to begin discussing their evaluation. This position paper presents the point of view that natural language interfaces offer special problems for evaluation, compared to problems encountered with other computer software. If ignored, such problems will make evaluation ineffectual. First, the paper describes the purposes and dimensions of problems in such an evaluation; differing purposes dictate differing approaches to evaluating natural language interfaces. Then, the paper points out some issues that turn out to be red herrings, and suggests a modest proposal that covers much of the broad spectrum of difficult issues involved. Since some of the issues and problems are common to both expert systems and natural language interfaces, comparisons of issues in evaluation of expert systems is also provided.

Based on the rapid development of Chinese agricultures, many English users are interesting on the Chinese agricultural information, and many Chinese users are interesting on English agricultural information too. This paper is a schema to design an agricultural cross languages engine, the core technology is the mapping between Chinese and English agricultural thesauri. The paper introduces the all rules of thesauri mapping, and give exact examples for these rules. With the mapping information, authors design a cross languages engine. English users can get Chinese agricultural information from web data by English descriptors; Chinese users can get English agricultural information from web data by Chinese descriptors

Computer-based chess machines have been built based on tree searches and parallel computing. Several papers have discussed the computer implementations of different types of chess games such as chess (western), Chinese chess (Xiangqi), Japanese chess (Shogi), and Hex. This paper focuses on the implementation of a triangle chess machine. We first make a brief explanation on the popular chess games and their related studies. We then describe the rules for playing a triangle chess. Basically, the triangle chess machine proposed in this paper considers all possible configurations of chess patterns and then makes moves based on these patterns. We present the software modules of the machine and discuss the computation issues related to this study. Conclusions and the directions for future studies are also given at the end of this paper.

Bayesian network is a powerful tool to represent and deal with uncertain knowledge. There exists much uncertainty in crop or animal disease. The construction of Bayesian network need much data and knowledge. But when data is scarce, some methods should be adopted to construct an effective Bayesian network. This paper introduces a disease diagnosis model based on Bayesian network, which is two-layered and obeys noisy-or assumption. Based on the two-layered structure, the relationship between nodes is obtained by domain knowledge. Based on the noisy-model, the conditional probability table is elicited by three methods, which are parameter learning, domain expert and the existing certainty factor model. In order to implement this model, a Bayesian network tool is developed. Finally, an example about cow disease diagnosis was implemented, which proved that the model discussed in this paper is an effective tool for some simple disease diagnosis in crop or animal field.

The Reference Model for Open Distributed Processing (RM-ODP) defines a framework within which support of distribution, interoperability, and portability can be integrated. However, other ODP standards have to be defined. We treat in this paper the need for formal notation for community related structural concepts in the enterprise language. Indeed, the ODP viewpoint languages are abstract in the sense that they define what concepts should be supported, not how these concepts should be represented. One approach to define the formal semantics of a language is denotational elaborating of the instance denoted by a sentence of the language in a particular context. Using the denotational semantics in the context of UML/OCL, we define in this paper semantics for the community related concepts defined in the RM-ODP foundations part and in the enterprise language. These specification concepts are suitable for describing and constraining ODP enterprise viewpoint specifications.

Many researchers and organizations are interested in creating a mechanism capable of automatically predicting software defects. In the last years, machine learning techniques have been used in several researches with this goal. Many recent researches use data originated from NASA (National Aeronautics and Space Administration) IV&V (Independent Verification & Validation) Facility Metrics Data Program (MDP). We have recently applied a constructive neural network (RBF-DDA) for this task, yet MLP neural networks were not investigated using these data. We have observed that these data sets contain inconsistent patterns, that is, patterns with the same input vector belonging to diAEerent classes. This paper has two main objectives, (i) to propose a modified version of RBF-DDA, named RBF-eDDA (RBF trained with enhanced Dynamic Decay Adjustment algorithm), which tackles inconsistent patterns, and (ii) to compare RBF-eDDA and MLP neural networks in software defects prediction. The simulations reported in this paper show that RBF-eDDA is able to correctly handle inconsistent patterns and that it obtains results comparable to those of MLP in the NASA data sets.

A web-based multi-models expert system called DCDDS is presented in this paper, which developed for diagnosis of dairy cow diseases through the symptoms submitted by users on web. As it is accepted that the inference engine and the relevant knowledge representation are the crucial part of diagnosis expert system, which limits its application and popularization in animal disease diagnosis. To break the limit and raise accuracy, this paper compares and appraises the existed systems and presents a solution that contains three models—Case-based reasoning (CBR), Subjective Bayesian theory and D-S evidential theory. Accordingly a knowledge representation method which can support the three different models is also designed. Up to the complicacy of the group of symptoms users acquired, they can choose which of the three models should be adopted to meet the best resolve. The performance of the proposed system was evaluated by an application to the field of dairy cow disease diagnosis using a real example of dairy cow diseases. The result indicates that the new methods have improved the inference procedures of the expert systems, and have showed that the new architecture has some advantage over the conventional architectures of expert systems on both efficiency and accuracy.

This paper describes design and creation of a multilingual parallel corpus for South African languages. One of the applications of the corpus, namely, the induction of a part-of-speech tagger for Afrikaans from the data, is presented in the paper. Development of the Afrikaans part-of-speech tagger is based on a modified method for induction of linguistic tools from parallel corpora originally proposed by Yarowsky and Ngai (2001).

This paper describes the development of a memory-based lemmatiser for Afrikaans called Lia. The paper commences with a brief overview of Afrikaans lemmatisation and it is indicated that lemmatisation is seen as a simplified process of morphological analysis within the context of this paper. This overview is followed by an introduction to memory-based learning — the machine learning technique that is used in the development of the Afrikaans lemmatiser. The deployment of Lia is then discussed with specific emphasis on the format of the training and testing data that is used. The Afrikaans lemmatiser is then evaluated and it is indicated that Lia achieves a linguistic accuracy figure of over 90%. The paper concludes with some ideas on future work that can be done to improve the linguistic accuracy of the Afrikaans lemmatiser.

This book contains a selection of the best papers given at the Advanced Computer Systems Conference in October 2006 in Poland. The book is partitioned into four topical areas: The Artificial Intelligence section includes papers on Humatronics, Fuzzy Logic, Robotics, Cell, Genetic and Probabilistic Algorithms, Incomplete Data and Similarity Relations, and Fast Neural Networks. The Computer Security and Safety section covers a wide range of issues, from Error Detection in Block Ciphers to Visual Cryptography Methods. The Image Analysis, Graphics and Biometrics section deals with the extraction of meaningful information from images as well as methods for uniquely recognizing humans based upon intrinsic physical or behavioral traits. The Computer Simulation and Data Analysis section concerns computer programs that attempt to simulate an abstract model of a particular system and transform data with the aim of extracting useful information and facilitating conclusions.

This paper introduces first an expert system shell based on stratified forward chaining. The stratified forward chaining was proposed as a generalization for the inheritance networks with exception since it allows more than one antecedent in the premises part of a rule. Expert systems built with this tool are rule based and the user interface is web based. The practical use of this chaining is presented here through the three chaining methods: forward chaining, backward chaining and mixed chaining. Our current aim is to make use of these inference mechanisms to help the user in his interaction with administration to identify in an efficient way the relevant information he needs. This is done through a question-response dialog to determine the user profile. The paper considers also in a second part a special kind of inference called deontic inference. The rationale behind this kind of inference is the possible presence of deontic rules which may impose a change in the base of current facts to make it comply with the obligations they prescribe.

In this paper the prototype of a system designed to analyze road accidents is presented. The analysis is carried out in order to recognize within accident reports general mechanisms of road accidents that represent prototypes of road accidents. Case Based Reasoning (CBR) is the chosen problem solving paradigm. Natural language documents and semi-structured documents are used to build the cases of our sys-tem, which creates a difficulty. To cope with this difficulty we propose approaches integrating semantic resources. Hence, an ontology of accidentology and a terminology of road accidents are used to build cases. The alignment of two resources supports the retrieval process. A data processing model, based on models of accidentology, is proposed to rep-resent the cases of the system. This paper presents the architecture of ACCTOS (ACCident TO Scenarios), a case based reasoning system prototype. The model to represent the cases is introduced and the phases of the case based reasoning cycle are detailed.

Fish disease diagnosis is a complicated process and requires high level of expertise, an expert system for fish disease diagnosis is considered as an effective tool to help fish farmers. However, many farmers have no computers and are not able to access the Internet. Telephone and mobile uses increase rapidly, so, the provision of call centre service appears as a sound alternative support channel for farmer to acquire counseling and support. This paper presents a research attempt to develop and evaluate a call center oriented Hybrid disease diagnosis & consulting system (H-Vet) in aquaculture in China. This paper looks at why H-Vet is needed and what are the advantages and difficulties in the developing and using such a system. A machine learning approach is adopted, which helps to acquire knowledge when enhancing expert systems with the user information collected through call center. This paper also proposes a fuzzy Group Support Systems (GSS) framework for acquiring knowledge from individual expert and aggregating knowledge into workgroup knowledge by H-Vet in the situation of difficult disease diagnosis. The system’s architecture and components are described.

This paper relates semantics as it is used in linguistics and natural language processing to the operational requirements of image retrieval systems. This is done in the context of a model of exploratory search and image annotation or indexing. The paper concludes this operational context requires the use of a restricted form of semantics compared with the usual one from linguistics or natural language processing, focussing on words rather sentences.

This paper suggests several estimation guidelines for the choice of a suitable machine learning technique for software development effort estimation. Initially, the paper presents a review of relevant published studies, pointing out pros and cons of specific machine learning methods. The techniques considered are Association Rules, Classification and Regression Trees, Bayesian Belief Networks, Neural Networks and Clustering, and they are compared in terms of accuracy, comprehensibility, applicability, causality and sensitivity. Finally the study proposes guidelines for choosing the appropriate technique, based on the size of the training data and the desirable features of the extracted estimation model.

International audience; This paper addresses the problem of automatic learning of scenarios. A ubiquitous computing environment must have the ability to perceive its occupants and their activities in order to recognize a context and to provide appropriate services. A context (a scenario) can be modeled as a temporal sequence of situations. Hard coding contexts by hand is a complex task. Our goal is to learn these context models based on a set of videos showing actors playing predefined scenarios. Once these models are learned, we can use them to classify new scenarios. Hidden Markov Models (HMMs) are particularly well suited for problems with a strong temporal structure; they are easily adaptable to variability of input and robust to noise. But two problems need to be addressed: how many HMMs do we need for all possible scenarios and how many states for each HMM. We propose in this paper an approach based on an incremental algorithm addressing these two problems. Under the best conditions we obtained the minimal error rate of 1.96% (2 errors in 102 validation entries).

In this paper, we discuss the visualization of multidimensional data. A well-known procedure for mapping data from a high-dimensional space onto a lower-dimensional one is Sammon’s mapping. The algorithm is oriented to minimize the projection error. We investigate an unsupervised backpropagation algorithm to train a multilayer feed-forward neural network (SAMANN) to perform the Sammon’s nonlinear projection. Sammon mapping has a disadvantage. It lacks generalization, which means that new points cannot be added to the obtained map without recalculating it. The SAMANN network offers the generalization ability of projecting new data, which is not present in the original Sammon’s projection algorithm. Retraining of the network when the new data points appear has been analyzed in this paper.

As a result of the impersonation of a user of a mobile terminal, sensitive information kept locally or accessible over the network can be abused. The means of masquerader detection are therefore needed to detect the cases of impersonation. In this paper, the problem of mobile-masquerader detection is considered as a problem of classifying the user behaviour as originating from the legitimate user or someone else. Different behavioural characteristics are analysed by designated one-class classifiers whose classifications are combined. The paper focuses on selecting the classifiers for mobile-masquerader detection. The selection process is conducted in two phases. First, the classification accuracies of classifiers are empirically evaluated, and inaccurate classifiers are excluded. After that, the accuracies of different classifier combinations are explored, and the combination with the best classification accuracy is identified. The experimental results suggest that, in order to achieve better accuracy, the individual classifiers with both high classification accuracy and a small number of non-classifications need to be selected.

One problem domain of meta search is to combine and improve the precision of ranking results from various search systems. This paper describes a rank aggregation model that incorporates text analysis measure with existing rank-based method, e.g. Best Rank and Borda Rank, to aggregate search results from various search systems. This approach provides means to normalize the differences of rank methodology used by different search systems, justifying the potential of using contents analysis to improve the results relevancy in meta search. In this paper, we fully describe our approach on text normalization for meta search and present our rationality of using two rank-based methods in our model. We then evaluate and benchmark the performance of our model based on user judgment on results relevancy. Our experiment results show that when text analysis factor is taken into account, the results outperform the rank-based methods alone. This shows the potential of our model to complement current rank aggregation methods used in meta search.

In this paper we present a novel data mining technique, known as Post Sequential Patterns Mining, which can be used to discover Structural Patterns. A Structural Pattern is a new pattern, which is composed of sequential patterns, branch patterns or iterative patterns. Sequential patterns mining plays an essential role in many areas and substantial research has been conducted on their analysis and applications. In our previous work [12], we used a simple but efficient Sequential Patterns Graph (SPG) to model the sequential patterns. The task to discover hidden Structural Pattern is based on our previous work and sequential patterns mining, conveniently named Post Sequential Patterns Mining. In this paper, in addition to stating this new mining problem, we define patterns such as branch pattern, iterative pattern, structural pattern, and concentrate on finding concurrent branch pattern. Concurrent branch pattern is thus one of the main forms of structural pattern and will play an important role in event-based data modelling.

The methodology of artificial evolution based on the traditional fitness function is argued to be inadequate for constructing the entities with behaviors novel to their designers. Evolutionary emergence via natural selection(without an explicit fitness function) is a promising way. This paper primarily considers the question of what to evolve, and focuses on the principles of developmental modularity based on neural networks. The connection weight values of this neural network are encoded as genes, and the fitness individuals are determined using a genetic algorithm. In paper we has created and described an artificial world containing autonomous organisms for developing and testing some novel ideas. Experimental results through simulation have demonstrated that the developmental system is well suited to long-term incremental evolution. Novel emergent strategies are identified both from an observer’s perspective and in terms of their neural mechanisms.

Reasoning Systems (Inference Mechanisms) and Neural Networks are two major areas of Artificial Intelligence (AI). The use of case-based reasoning in Artificial Intelligence systems is well known. Similarly, the AI literature is replete with papers on neural networks. However, there is relatively little research in which the theories of case-based reasoning and neural networks are combined. In this paper we integrate the two theories and show how the resulting model is used in a medical diagnosis application. An implementation of our model provides a valuable prototype for medical experts and medical students alike.

In this paper we address the problem of training multilayer feed-forward neural networks. These networks have been widely used for both prediction and classification in many different areas. Although the most popular method for training these networks is back propagation, other optimization methods such as tabu search or scatter search have been applied to solve this problem. This paper presents a new training algorithm based on the tabu search methodology that incorporates elements for search intensification and diversification by utilizing strategic designs where other previous approaches resort to randomization. Our method considers context and search information, as it is provided by the partial derivatives and memory structures, for move selection. The experimentation shows that the proposed procedure can compete with the best-known algorithms in terms of solution quality, consuming a reasonable computational effort.

This paper describes new approaches to classification/identification of biological data. It is expected that the work: may be extensible to other domains such as the medical domain or fault diagnostic problems. Organisms are often classified according to the value of tests which are used for measuring some characteristic of the organism. When selecting a suitable test set it is important to choose one of minimum cost. Equally, when classification models are constructed for the posterior identification of unnamed individuals it is important to produce optimal models in terms of identification performance and cost. In this paper, we first describe the problem of selecting an economic test set for classification. We develop a criterion for differentiation of organisms which may encompass fuzzy differentiability. Then, we describe the problem of using batches of tests sequentially for identification of unknown organisms, and we explore the problem of constructing the best sequence of batches of tests in terms of cost and identification performance. We discuss how metaheuristic algorithms may be used in the solution of these problems. We also present an application of the above to the problem of yeast classification and identification.

Despite UML being considered a software engineering standard, the UML syntactic notations used in texts, papers, documentation and CASE tools are often different. The decision as to which of the semantically equivalent notational variations to use appears to be according to the personal preference of the author or publisher, rather than based on any consideration of the ease with which the notation can be understood by human readers. This paper reports on an experiment that takes a human comprehension perspective on UML class diagram notational variants. Five notations were considered: for each, two semantically equivalent, yet syntactically different, variations were chosen from published texts. Our experiment required subjects to indicate whether a supplied specification matched each of a set of experimental diagrams. The results reveal that the best performing notation may depend on the task for which it is used, and that our personal, intuitive predictions intuitions (which were based in the complexity of the notation) were partly confirmed.

The topic of this paper is the design and implementation of an interacting Lego digger and dumper truck through the use of Petri nets. The focus of this is primarily on the use of Petri nets in the developing of dependable systems, an area of particular concern in the full-size equivalent of our experiment. The content of this paper will discuss the progression of the Petri net model of these two Lego robots, from design to implementation, and finally evaluation. The paper also features an optional plug-in Petri net, intended to allow the dumper to authenticate the digger for added security.

This paper describes the computer aided diagnosis of breast cancer using three dimensional ultrasonic echography. The authors have already proposed the algorithm for automated breast tumor extraction from ultrasonic echographic images. 1,2,3 Three dimensional voxel data is constructed from a series of the cross sections is captured from video signals and probe position and orientation data which are simultaneously obtained by 3D space tracking system. In the algorithm the tumor is extracted by using the following concepts; voxel in tumor region is low echo level, each voxel is classified as one of “tumor”, “normal tissue”, or “boundary” by using fuzzy reasoning and membership function is calculated from outputs of 3D LoG filtering. In this paper we propose the improved algorithm with multiple filters thus we obtain over 90% successfulness for good agreement of computed contour with manual traced contour. It is important to judge the agreement of tumor contour by computer because various kinds of filters are iteratively applied the voxel data in the proposed algorithm. Therefore we try to judge the agreement of contour by fuzzy reasoning. The basic concepts for fuzzy reasoning are that voxel inside the tumor region is low echo level and the morphology of the tumor is reasonable. For over ninety percent of extracted tumor good agreements of computed judge with human judge are found. The results say that the proposed algorithm has a potential for computer aided diagnosis of breast cancer.

Classification of patient samples is a crucial aspect of cancer diagnosis. DNA hybridization arrays simultaneously measure the expression levels of thousands of genes and it has been suggested that gene expression may provide the additional information needed to improve cancer classification and diagnosis. This paper presents methods for analyzing gene expression data to classify cancer types. Machine learning techniques, such as Bayesian networks, neural trees, and radial basis function (RBF) networks, are used for the analysis of the CAMDA Data Set 2. These techniques have their own properties including the ability of finding important genes for cancer classification, revealing relationships among genes, and classifying cancer. This paper reports on comparative evaluation of the experimental results of these methods.

Real-time segmentation of moving regions in image sequences is a fundamental step in many vision systems including automated visual surveillance, human-machine interface, and very low-bandwidth telecommunications. A typical method is background subtraction. Many background models have been introduced to deal with different problems. One of the successful solutions to these problems is to use a multi-colour background model per pixel proposed by Grimson et al [1, 2,3]. However, the method suffers from slow learning at the beginning, especially in busy environments. In addition, it can not distinguish between moving shadows and moving objects. This paper presents a method which improves this adaptive background mixture model. By reinvestigating the update equations, we utilise different equations at different phases. This allows our system learn faster and more accurately as well as adapts effectively to changing environment. A shadow detection scheme is also introduced in this paper. It is based on a computational colour space that makes use of our background model. A comparison has been made between the two algorithms. The results show the speed of learning and the accuracy of the model using our update algorithm over the Grimson et al’s tracker. When incorporate with the shadow detection, our method results in far better segmentation than The Thirteenth Conference on Uncertainty in Artificial Intelligence that of Grimson et al.

Existing classification and rule learning algorithms in machine learning mainly use heuristic/greedy search to find a subset of regularities (e.g., a decision tree or a set of rules) in data for classification. In the past few years, extensive research was done in the database community on learning rules using exhaustive search under the name of association rule mining. The objective there is to find all rules in data that satisfy the user-specified minimum support and minimum confidence. Although the whole set of rules may not be used directly for accurate classification, effective and efficient classifiers have been built using the rules. This paper aims to improve such an exhaustive search based classification system CBA (Classification Based on Associations). The main strength of this system is that it is able to use the most accurate rules for classification. However, it also has weaknesses. This paper proposes two new techniques to deal with these weaknesses. This results in remarkably accurate classifiers. Experiments on a set of 34 benchmark datasets show that on average the new techniques reduce the error of CBA by 17% and is superior to CBA on 26 of the 34 datasets. They reduce the error of the decision tree classifier C4.5 by 19%, and improve performance on 29 datasets. Similar good results are also achieved against the existing classification systems, RIPPER, LB and a Naive-Bayes classifier.

System-level fault diagnosis of massively parallel computers requires efficient algorithms, handling a many processing elements in a heterogeneous environment. Probabilistic fault diagnosis is an approach to make the diagnostic problem both easier to solve and more generally applicable. The price to pay for these advantages is that the diagnostic result is no longer guaranteed to be correct and complete in every fault situation. In an earlier paper [2] the authors presented a novel methodology, called local information diagnosis, and applied it to create a family of probabilistic diagnostic algorithms. This paper examines the identification of fault-free and faulty units in detail by defining three heuristic methods of fault classification and comparing the diagnostic accuracy provided by these heuristics using measurement results.

Refinement tools seek to correct faulty rule-based systems by identifying and repairing faults that are indicated by training examples that provide some evidence of faults. Refinement tools typically use a hill-climbing search to identify suitable repairs. In this paper, the goal is to incorporate an effective backtracking mechanism with a refinement algorithm so that the search for repairs does not get caught by local maxima. However the repair cycle for each potential fault is expensive, so exhaustive backtracking is prohibitive for large knowledge bases. This paper investigates more guided backtracking algorithms developed for constraint satisfaction problems and adapts them for refinement problems. Experiments with these backtracking algorithms reveal that high accuracy refined knowledge bases are achievable, often at the expense of extra iterations, but an informed re-ordering of training examples reduces the number of iterations without increasing the error-rate. A test-bed is developed by corrupting a rule base with interacting faults, thereby allowing pairs of conflicting training examples to be identified. The algorithms are evaluated on training sets containing increasing numbers of these conflicting examples. One separate observation is that conflicting examples help to achieve refined knowledge bases with high accuracy.

The purpose of this paper is to explore the potential value of applying language-based methods of analyses to textual ethnographic field records (like e-mail records) in addition to thick descriptions. Even though language analysis imports many a priori assumptions, they could provide more concise summaries than thick descriptions and add value to them. The methods of conversational analysis investigated here a based on a coding scheme derived from the Wittgenstein “language use” theory of meaning combined with Habermas’ social action typology. The paper presents some coding issues fordiscussion that we encountered along with illustrative tabulations of sample coding results. Our purpose is to illustrate the potential value of language action coding to ethnographers both during the analysis and reporting stages of research.

This short overview paper provides a preliminary investigation of the potentials of possibilistic logic in decision analysis. Indeed a possibilistic logic base can not only be seen as a set of more or less certain pieces of information (which was the original understanding when possibilistic logic was introduced), but also as a layered set of propositions expressing goal shaving different levels of priority. The paper surveys applications to multiple criteria decision (specification of preferences and goals, modeling of various types of aggregation and weighting procedures), and to decision under uncertainty. Preference revision or combination, and analysis of conflicts between goals are also briefly discussed. The possibilistic logic framework clearly supports a qualitative view of decision based on the use of ordinal scales.

In recent years, there has been an increasing number of papers in the literature that use neural networks for financial time-series prediction. This paper focuses on the Greek Foreign Exchange Rate Currency Market. Using four major currencies, namely the U.S. Dollar (USD), the Deutsche Mark (DM), the French Franc (FF) and the British Pound (BP), against the Greek Drachma, on a daily basis for a period of 11 years, we try to forecast the above time-series, using two different approaches. The first one involves RBF training using a Kalman filter, while the second one uses genetic algorithms in order to optimize the RBF network parameters. The goal of this effort, is, first, to predict, as accurately as possible, currencies future behaviour, in a daily predicting horizon and second, to have a comparison between the two methods.

This paper describes an ongoing work on the application of machine learning techniques in the domain of water distribution networks. This research is being done in the context of the European Esprit project Waternet. One part of this project is a learning system which intends to capture knowledge from historic information collected during the operation of a water distribution network. Captured knowledge is expected to contribute to improve the operation of the network. The ideas presented in this paper describe the first development phase of this learning system, focusing specially in the practical methodology adopted. The interaction between different classes of human experts and the learning system are discussed Finally some preliminary experimental results are presented.

Inheritance is a powerful concept that has been extensively used in the context of data modeling in databases, programming in object-oriented paradigm, and knowledge representation in artificial intelligence. Because of the close relationship between inheritance and deduction, logic is one of the obvious candidates for a declarative framework for inheritance. However, inheritance has resisted many serious attempts at a satisfactory logical rendition. In this paper, we shall survey various proposals in the literature for formalizations of inheritance and classify them based on the kind of inheritance they capture (structural or behavioral, monotonic or nonmonotonic, value inheritance or definition inheritance), and also based on their expressive power. We also classify the proposals from the perspective of the underlying logical framework they are based on (e.g., proof theory or model theory, etc). We conclude the paper with a list of what, from our perspective, are the most significant problems requiring further research in this field. Our focus in this paper will be primarily on the database and knowledge representation fronts.

All types of automated manufacturing systems, which are often balanced automation systems are broadly using robots, manipulators, etc. This paper gives a restricted survey on the application of some of the so called artificial intelligence (AI) methods and tools which can be used to model, simulate, design and implement systems with robot and manipulator applications. Al in design and planning of robotic systems means the application of Multi-Agent Systems (MAS), Expert Systems (ES), Knowledge Based Systems (KB S), Artificial Neural Networks (ANN), Petri nets, Fuzzy Systems (FS), Genetic Algorithms (GA), Object-Oriented Technologies (OO), and several other things including pattern recognition, vision, voice and tactile systems and learning. Modelling and simulation in our recent understanding are more design assistance tools, than the goal of AI applications thus they completely fit into our recent topics. The paper deals not only with manufacturing and assembly applications, but with other intelligent robotic issues as well, as autonomous robots, moving robots, surgery robots, etc.

In order to acquire knowledge from databases, there have been proposed several methods of inductive learning, such as ID3 family and AQ family. These methods are applied to discover meaningful knowledge from large databases, which shows they are useful. However, since there has been no formal approach proposed to treat these methods, efficiency of each method is only compared empirically. In this paper, we introduce matroid theory and rough sets to construct a common framework for empirical machine learning methods which induce the combination of attribute-value pairs from databases. Combination of the concepts of rough sets and matroid theory gives us an excellent set-theoretical framework and enables us to understand the differences and the similarities between these methods clearly. In this paper, we compare three classical methods, AQ, Pawlak’s Consistent Rules and ID3. The results show that there exist the differences in algebraic structure between the former two and the latter and that this causes the differences between AQ and ID3.

This paper discusses problem solving and its algorithmic-like properties. In systems literature problem solving is often discussed in relation to its methodological setting — for example, SSM may legitimately be regarded as a problem solving scheme. It is a contention of this paper that five cognitive elemental strategies may be involved in generalised problem solving.

This paper presents applications of graph theory to the design of graph matching neural networks for automated fingerprint identification. Given a sparse set of minutiae from a fingerprint image, complete with locations in the plane and (optionally) other labels such as ridge angles, ridge counts to nearby minutiae and so on, this approach to matching begins by constructing a graph-like representation of the minutiae map, utilizing proximity graphs, such as the sphere-of-influence graphs. These graph representations are more robust to noise such as translations, rotations and deformations. This paper presents the role of these graph representations in the design of graph matching neural networks for the matching and classification of fingerprint images.

In this paper a generic framework inspired by [LX90] is presented for the validation of reactive systems embedded in a test environment, or isolated from their operational environment, thereby inducing a natural classification of validation strategies in different scenarios. We show that verification strategies and falsification strategies are complementary notions (but not necessarily equally strong), and how the choice of strategy for assessing the correctness of systems in different scenarios affects the outcome of the validation process. The significance of this paper lies in the provision of systematic principles to perform validation in context, and it gives a theoretical explanation of validation strategies that have already been applied informally in practice. The theory in this paper provides a guide for performing correctness assessment of systems in context, and shows that the correctness of systems in such cases can be approximated in the sense that the risk of considering an incorrect system correct can be reduced systematically.

Case-based reasoning is reasoning based on specific instances of past experience. A new solution is generated by retrieving and adapting an old one which approximately matches the current situation. In this paper, we outline a case-based reasoning scheme for path planning in autonomous underwater vehicle (AUV) missions. An annotated map database is employed to model the navigational environment. Routes which are used in earlier missions are represented as objects in the map. When a new route is to be planned, the path planner retrieves a matching route from the database and modifies it to suit to the current situation. Whenever a matching route is not available, a new route is synthesized based on past cases that describe similar navigational environments. Case-based approach is thus used not only to adapt old routes but also to synthesize new ones. Since the proposed scheme is centered around reuse of old routes, it would be fast especially when long routes need to be generated. Moreover, better reliability of paths can be expected as they are adapted from earlier missions. The scheme is novel and appropriate for AUV mission scenarios. In this paper, we describe the representation of navigation environment including past routes and objects in the navigational space. Further, we discuss the retrieval and repair strategies and the scheme for synthesizing new routes. Sample results of both synthesis and reuse of routes and system performance analysis are also presented. One major advantage of this system is the facility to enrich the map database with new routes as they are generated.

The paper addresses an important issue: how to involve the users of a system into the development process of the system. Since different types of software systems impose different requirements on the system development process the paper confines itself to the development of one class of systems: data-intensive dialogue systems.

Handwritten digit and alphabetic character recognition has been an active area of research for many years. Character recognition techniques associate a symbolic identity with the image of a character. The problem of replication of human functions by machines involves the recognition of both machine and hand-printed or cursive-written characters. Character recognition is better known as Optical Character Recognition (OCR) since it deals with recognition of optically processed characters rather than, for example, magnetically processed ones. The origin of character recognition can be found as early as in 1870, though the modern version of OCR appeared in the middle of the 1940s with the development of digital computers. Its main area of interest was data processing with application to the business world. The principal motivation for development of OCR system is the need to cope with the enormous flood of paper such as bank cheques, commercial forms, government records, mail sorting generated by the expanding society and international partnerships. Since the origin of OCR systems extensive research has been carried out and a large number of technical papers have been published by researchers in this area. Many techniques and experimental results have been published in the literature. In recent years, new classification approaches, such as neural networks, and design methodologies, such as classifier combination and parallel feature networks extraction, have renewed interests in this fields. As OCR technology seeks applications in more complicated, real time environments, the criterion for successful algorithm becomes more rigorous. New feature extraction and classification techniques are developed with parallel hardware implementation in mind. Several recognition methodologies are combined in order to improve the recognition quality. State of the art reports on character recognition research have been presented by Nagy [1968], Harmon [1972], Suen et al. [1980], Mori et al. [1984], Mantas [1986], Davis and Yall [1986], Govindan and Shivaprasad [1990], Lee and Srihari [1993].

Most existing work on concept formation (e.g., conceptual clustering) has focused on the formation of concepts for classification purposes, resulting in an emphasis on producing sets of concepts that are complete, i.e., cover all observed instances. In this research, we are primarily concerned with concept formation as a solution to the new term problem, which results in a new set of requirements on concept formation: concept formation must be triggered in a demand-driven fashion by the shortcomings of the existing representation, and a concept's quality must be evaluated in terms of its contribution to the representation. The approach to demand-driven concept formation that we present in this paper is part of an integrated learning system called the MODELER, the learning component of the knowledge acquisition system BLIP, and is capable of forming concepts incorporating both disjunctions and relational information. Much of the power of the approach stems from its close integration with the other components of the learning system, namely, rule discovery and knowledge revision, which are also presented in this paper.

This paper presents the development of a hydrogeological expert system able to handle the drilling location problem within the scope of village water-supply programs. This work is based on the experience the authors gained from thousands of drillings carried out in fifteen African countries. The cognitive model comes from the practical know-how acquired from real-world programs, from original statistics and probabilistics analyses showing connections between data collected during the drilling-site selection and hydrodynamic parameters registered in the borings and from the research the authors carried out in the artificial intelligence field to propose a comprehensive knowledge modeling framework. The paper includes a description of the specific knowledge involved in the drilling location process. Relevant hydrogeological parameters recognition and examples of advanced computer knowledge modeling methods are presented. First the rules of thumb and the interpretative frames retained for the cognitive model are described, then the characteristics of the HYDROLAB expert system devoted both to computer-aided decision and to computer-assisted learning support.

A novel work of reference is proposed which is to be implemented as a writing aid in the form of software tools. This reference source is named Chusaurus and is based on a compilation of well-written, useful expressions extracted from papers and books which could be used to assist in the generation of scientific text. The expressions have blank spaces (representing missing words) which should be filled in by the user to suit his/her needs. The most important feature of the Chusaurus lies in the help it provides to non-native users of English in coping very rapidly with linguistic difficulties, since it provides pieces of real language in use. For example, it assists them in producing correct phase structuring with adequate vocabulary, which is one of the aspects of English that can be extremely different from their own native language. Two possible implementations are discussed. In the first, the resourse could be implemented as an extra function in word processors.The expressions would be accessed from menus with the compiled expressions being indexed according to their suitability for use in the different sections (and for the subjects) most likely to appear in a paper. The second implementation is more sophisticated and involves the development of a writing environment in which support tools would assist the user in selecting and organizing the material to be written as well as in producing a well-written text.

This paper discusses the planning of multisentential text by computer. In order to construct coherent paragraphs, we have been using relations from Rhetorical Structure Theory (RST) operationalized as plans. The paper first describes, in some detail, the current method of planning a paragraph using operationalized RST relation/plans. It then makes two points that illustrate why RST relation/plans are an ideal tool for planning paragraphs. First, these relation/plans can be shown to combine the best features of paragraph-sized schemas and clause-sized planning rules under a top-down planning regime in a way which affords much flexibility to the user. Second, RST relation/plans can support both standard top-down planning and open-ended conversation-like behavior; a small difference in treatment gives rise to either paradigm.

This paper discusses the problem of lexical choice in text generation, suggesting that it has to be viewed as part of the problem of lexicogrammatical choice instead of being isolated. First, the importance of the problem and the need for a general survey are noted. Next, in order to interpret the contributions these approaches have made and to identify the full spectrum of factors that have to be considered, an abstract model of lexis within lexicogrammar is presented. Then, existing approaches are discussed and located in relation to the abstract model. This paper is a shorter version of a longer treatment that also goes into the abstract model presented here in more detail.