International Journal on Artificial Intelligence Tools

The articles in this special issue represent advances in several areas of knowledge acquisition and knowledge representation. In this article we attempt to place these advances in the context of a fundamental challenge in AI; namely, the automated acquisition of knowledge from data and the representation of this knowledge to support understanding and reasoning. We observe that while this work does indeed advance the field in important areas, the need exists to integrate these components into an end-to-end system and begin to extract general methodologies for this challenge. At the heart of this integration is the need for performance feedback throughout the process to guide the selection of alternative methods, the support for human interaction in the process, and the definition of general metrics and testbeds to evaluate progress.

Traditional host-based anomaly detection systems model normal behavior of applications by analyzing system call sequences. The current sequence is then examined (using the model) for anomalous behavior, which could correspond to attacks. Though these techniques have been shown to be quite effective, a key element is missing – the inclusion and utilization of the system call arguments. Recent research shows that sequence-based systems are prone to evasion. We propose an idea of learning different representations for system call arguments. Results indicate that this information can be effectively used for detecting more attacks than traditional sequence-based techniques, with reasonable storage and computational overhead.

We present an efficient feature selection algorithm for computer aided detection (CAD) computed tomographic (CT) colonography. The algorithm (1) determines an appropriate piecewise linear network (PLN) model by cross validation, (2) applies the orthonormal least square (OLS) procedure to the PLN model utilizing a Modified Schmidt procedure, and (3) uses a floating search algorithm to select features that minimize the output variance. The undesirable "nesting effect" is prevented by the floating search approach, and the piecewise linear OLS procedure makes this algorithm very computationally efficient because the Modified Schmidt procedure only requires one data pass during the whole searching process. The selected features are compared to those obtained by other methods, through cross validation with support vector machines (SVMs).

Historically, data mining research has been focused on discovering sets of attributes that discriminate data entities into classes or association rules between attributes. In contrast, we are working to develop data mining techniques to discover patterns consisting of complex relationships between entities. Our research is particularly applicable to domains in which the data is event driven, such as counter-terrorism intelligence analysis. In this paper we describe an algorithm designed to operate over relational data received from a continuous stream. Our approach includes a mechanism for summarizing discoveries from previous data increments so that the globally best patterns can be computed by examining only the new data increment. We then describe a method by which relational dependencies that span across temporal increment boundaries can be efficiently resolved so that additional pattern instances, which do not reside entirely in a single data increment, can be discovered. We also describe a method for change detection using a measure of central tendency designed for graph data. We contrast two formulations of the change detection process and demonstrate the ability to identify salient changes along meaningful dimensions and recognize trends in a relational data stream.

Increasing availability of information has furthered the need for recommender systems across a variety of domains. These systems are designed to tailor each user's information space to suit their particular information needs. Collaborative filtering is a successful and popular technique for producing recommendations based on similarities in users' tastes and opinions. Our work focusses on these similarities and the fact that current techniques for defining which users contribute to recommendation are in need of improvement.

In this paper we propose the use of trustworthiness as an improvement to this situation. In particular, we define and empirically test a technique for eliciting trust values for each producer of a recommendation based on that user's history of contributions to recommendations.

We compute a recommendation range to present to a target user. This is done by leveraging under/overestimate errors in users' past contributions in the recommendation process. We present three different models to compute this range. Our evaluation shows how this trust-based technique can be easily incorporated into a standard collaborative filtering algorithm and we define a fair comparison in which our technique outperforms a benchmark algorithm in predictive accuracy.

We aim to show that the presentation of absolute rating predictions to users is more likely to reduce user trust in the recommendation system than presentation of a range of rating predictions. To evaluate the trust benefits resulting from the transparency of our recommendation range techniques, we carry out user-satisfaction trials on BoozerChoozer, a pub recommendation system. Our user-satisfaction results show that the recommendation range techniques perform up to twice as well as the benchmark.

In this paper we extend the cultural framework previously developed for the Village multi-agent simulation in Swarm to include the emergence of a hub network from two base networks. The first base network is kinship, over which generalized reciprocal exchange is defined, and the second is the economic network where agents carry out balanced reciprocal exchange. Agents, or households, are able to procure several resources. We use Cultural Algorithms as a framework for the emergence of social intelligence at both individual and cultural levels. Successful agents in both networks can promote themselves to be included in the hub network where they can develop exchange links to other hubs. The collective effect of the hub network is representative of the quality of life in the population and serves as an indicator for motives behind the mysterious emigration from the region. Knowledge represents the development and use of exchange relationships between agents. The presence of defectors in the hub network improved resilience of the social system while maintaining the population size at that observed where no defectors were present.

We have created a logic-based, Turing-complete language for stochastic modeling. Since the inference scheme for this language is based on a variant of Pearl's loopy belief propagation algorithm, we call it Loopy Logic. Traditional Bayesian networks have limited expressive power, basically constrained to finite domains as in the propositional calculus. Our language contains variables that can capture general classes of situations, events and relationships. A first-order language is also able to reason about potentially infinite classes and situations using constructs such as hidden Markov models(HMMs). Our language uses an Expectation-Maximization (EM) type learning of parameters. This has a natural fit with the Loopy Belief Propagation used for inference since both can be viewed as iterative message passing algorithms. We present the syntax and theoretical foundations for our Loopy Logic language. We then demonstrate three examples of stochastic modeling and diagnosis that explore the representational power of the language. A mechanical fault detection example displays how Loopy Logic can model time-series processes using an HMM variant. A digital circuit example exhibits the probabilistic modeling capabilities, and finally, a parameter fitting example demonstrates the power for learning unknown stochastic values.

Despite extensive theoretical work carried out on coordination during the last three decades, much research is still devoted to proving the merits of one model over another. In our article, we will show how the coordination process in French (with using the conjunction et (and)) can be explained through Applicative and Combinatory Categorial Grammar (ACCG).

Natural-language based knowledge representations borrow their expressiveness from the semantics of language. One such knowledge representation technique is Latent semantic analysis (LSA), a statistical, corpus-based method for representing knowledge. It has been successfully used in a variety of applications including intelligent tutoring systems, essay grading and coherence metrics. The advantage of LSA is that it is efficient in representing world knowledge without the need for manual coding of relations and that it has in fact been considered to simulate aspects of human knowledge representation. An overview of LSA applications will be given, followed by some further explorations of the use of LSA. These explorations focus on the idea that the power of LSA can be amplified by considering semantic fields of text units instead of pairs of text units. Examples are given for semantic networks, category membership, typicality, spatiality and temporality, showing new evidence for LSA as a mechanism for knowledge representation. The results of such tests show that while the mechanism behind LSA is unique, it is flexible enough to replicate results in different corpora and languages.

This article constitutes a contribution to an analysis of the notion of variable. Within the framework of Combinatory Logic as a formalism without bound variables, the Logic of Determination of Objects (LDO) provides an explanation for the necessary distinction between "whatever, any" and "indeterminate, indefinite" used by the introduction and elimination rules of quantifiers in Natural Deduction. The intension of a concept and typical and atypical occurrences of a concept are also introduced yielding new quantifiers which are more adequate to natural language processing (NPL) and to the study of natural inferences in common reasoning.

Automated Theorem Proving (ATP) systems are complex pieces of software, and thus may have bugs that make them unsound. In order to guard against unsoundness, the derivations output by an ATP system may be semantically verified by trusted ATP systems that check the required semantic properties of each inference step. Such verification needs to be augmented by structural verification that checks that inferences have been used correctly in the context of the overall derivation. This paper describes techniques for semantic verification of derivations, and reports on their implementation and testing in the GDV verifier.

In the past decade, ontology has been actively researched in various domains. The different ontological tasks range from simple language modeling in the linguistic domain, to semantic integration in the semantic web, and recently to application-specific tasks such as financial fraud management. We follow the trend and attempt to tackle some of the ontological problems in security management. The most complicated problem out of all is the semantic interoperability problem that is evident in the existence of various types of security elements such as IDS, firewall and virus scanner. Another problem is related to the semantic modeling tasks required for autonomous and intelligent reasoning. Semantic modeling of security events is essential for automatic and intelligent event correlation tasks that analyze semantically the different sources of security information to more accurately present the holistic network security status. We present in this paper a novel and formal ontology mapping approach and security ontology for the supporting and possibly resolution of these problems.

In this paper, we present a method with a low complexity for analysing short and dynamic biomedical sequences. The method uses the Daubechies D4 wavelet in combination with similarity fitness schemes for retrieval. We also present a more traditional way of retrieving dynamical biomedical sequences based on stretching the sequences in time. The first mentioned method has been shown to outperform Fourier based methods in retrieving biomedical sequences of dynamic lengths, as well as the Haar wavelet.