Data mining techniques for extracting knowledge from text have been applied extensively to applications including question answering, document summarisation, event extraction and trend monitoring. However, current methods have mainly been tested on small-scale customised data sets for specific purposes. The availability of large volumes of data and high-velocity data streams (such as social media feeds) motivates the need to automatically extract knowledge from such data sources and to generalise existing approaches to more practical applications. Recently, several architectures have been proposed for what we call knowledge mining: integrating data mining for knowledge extraction from unstructured text (possibly making use of a knowledge base), and at the same time, consistently incorporating this new information into the knowledge base. After describing a number of existing knowledge mining systems, we review the state-of-the-art literature on both current text mining methods (emphasising stream mining) and techniques for the construction and maintenance of knowledge bases. In particular, we focus on mining entities and relations from unstructured text data sources, entity disambiguation, entity linking and question answering. We conclude by highlighting general trends in knowledge mining research and identifying problems that require further research to enable more extensive use of knowledge bases.

This paper provides a survey of the state of the art in plausible reasoning, that is exception tolerant reasoning under incomplete information. Three requirements are necessary for a formalism in order to cope with this problem: (i) making a clear distinction between factual information and generic knowledge; (ii) having a correct representation of partial ignorance; (iii) providing a nonmonotonic inference mechanism. Classical logic fails on requirements (i) and (iii), whilst the Bayesian approach does not fulfil (ii) in an unbiased way. In this perspective, various uncertainty modelling frameworks are reviewed: MYCIN-like fully compositional calculi, belief functions, upper and lower probability systems, and possibility theory. Possibility theory enables classical logic to be extended to layered sets of formulae, where layers express certainty levels. Finally, it is explained how generic knowledge can be expressed by constraints on possibility measures, and how possibilistic inferences can encode nonmonotonic reasoning in agreement with the Lehmann et al. postulates.

Despite the historical difference in focus between AI planning techniques and Integer Programming (IP) techniques, recent research has shown that IP techniques show significant promise in their ability to solve AI planning problems. This paper provides approaches to encode AI planning problems as IP problems, describes some of the more significant issues that arise in using IP for AI planning, and discusses promising directions for future research.

Network providers either attempt to handle massive distributed denial-of-service attacks themselves or redirect traffic to third-party scrubbing centers. If providers adopt the first option, it is sensible to counter such attacks in their infancy via provider collaborations deploying distributed security mechanisms across multiple domains in an attack path. This motivated our work presented in this paper. Specifically, we investigate the establishment of trusted federations among adjacent and disjoint network domains, that is, autonomous systems (ASes) that collectively mitigate malicious traffic. Our approach is based on Distributed Ledger Technologies for signaling, coordination, and orchestration of a collaborative mitigation schema via appropriate blockchain-based smart contracts. Reputation scores are used to rank ASes based on their mitigation track record. The allocation of defense resources across multiple collaborators is modeled as a combinatorial optimization problem considering reputation scores and network flow weights. Malicious flows are mitigated using programmable network data paths within the eXpress Data Path (XDP) framework; this enables operators with enhanced packet processing throughput and advanced filtering flexibility. Our schema was implemented in a proof-of-concept prototype and tested under realistic network conditions.

Research into visual reasoning up to now has focused on images that are literal depictions of their objects. I argue in this article that an important further mode of visual reasoning operates on images that depict objects metaphorically. Such images form part of the class of expressive symbols: they are found, for example, in allegorical representations in works of visual art, studied by iconology. They were also a common way of encapsulating insights about the universe in natural philosophy in the Renaissance. Many writers assume that expressive symbols have vanished from modern science, but I argue in the second part of the article that mathematical law statements in present-day physics should be seen, in part, as images that constitute expressive symbols of the world. In support of this view, I offer evidence that law statements relate to their objects metaphorically and that physicists engage with them primarily through visual inspection and visual reasoning.

Knowledge engineering (KE) has now provided some effective techniques for formalization of knowledge about goals and actions. These techniques could open new areas of research to statisticians. Experimental systems designed to assist users of statistics have been constructed in experiment design, data analysis, technique application, and technique selection. Knowledge formalization has also been used in experimental programs to assist statisticians in doing data analysis and in building consultation systems. The best-explored application of KE techniques is building consultation systems. It is now a promising area for development. Analogies with successful artificial intelligence AI applications in other fields suggest other statistical applications worth exploring. Opening new areas to research and providing new tools to users would make considerable changes in the use and production of statistical techniques. However, applying currently available KE techniques will lead to more work for statisticians, not less.

This paper presents our preliminary research into the autonomous control of an alpine skiing robot. Based on our previous experience with active balancing on difficult terrain and developing an ice-skating robot, we have implemented a simple control system that allows the humanoid robot Jennifer to steer around a simple alpine skiing course, brake, and actively control the pitch and roll of the skis in order to maintain stability on hills with variable inclination.

The robot steers and brakes by using the edges of the skis to dig into the snow, by inclining both skis to one side the robot can turn in an arc. By rolling the skis outward and pointing the toes together the robot creates a snowplough shape that rapidly reduces its forward velocity.

To keep the skis in constant contact with the hill we use two independent proportional-integral-derivative (PID) controllers to continually adjust the robot’s inclination in the frontal and sagittal planes.

Our experiments show that these techniques are sufficient to allow a small humanoid robot to alpine ski autonomously down hills of different inclination with variable snow conditions.

The traditions of image processing and knowledge engineering have developed separately. Work on AI vision systems lies between the two traditions but only recently has attention been given to combining practical imaging systems with methods for exploiting knowledge in interpreting the contents of an image. Five general approaches to combining knowledge based expert systems with imaging technologies are discussed. Particular attention is paid to the requirement for techniques which transform a pixel array into a symbolic form suitable for interpretation, and current obstacles to a general solution. Interpretation of biomedical images is particularly problematic because of statistical, structural and temporal variation in morphology of objects and structures. Some ways in which knowledge of shape, structure, and object classifications may contribute to this interpretation are discussed. The survey focuses on biomedical images but many of the issues are of general relevance to work in image understanding.

We consider task planning for long-living intelligent agents situated in dynamic environments. Specifically, we address the problem of incomplete knowledge of the world due to the addition of new objects with unknown action models. We propose a multilayered agent architecture that uses meta-reasoning to control hierarchical task planning and situated learning, monitor expectations generated by a plan against world observations, forms goals and rewards for the situated reinforcement learner, and learns the missing planning knowledge relevant to the new objects. We use occupancy grids as a low-level representation for the high-level expectations to capture changes in the physical world due to the additional objects, and provide a similarity method for detecting discrepancies between the expectations and the observations at run-time; the meta-reasoner uses these discrepancies to formulate goals and rewards for the learner, and the learned policies are added to the hierarchical task network plan library for future re-use. We describe our experiments in the Minecraft and Gazebo microworlds to demonstrate the efficacy of the architecture and the technique for learning. We test our approach against an ablated reinforcement learning (RL) version, and our results indicate this form of expectation enhances the learning curve for RL while being more generic than propositional representations.

The terminology of Machine Learning and Data Mining methods does not always allow a simple match between practical problems and methods. While some problems look similar from the user's point of view, but require different methods to be solved, some others look very different, yet they can be solved by applying the same methods and tools. Choosing appropriate Machine Learning methods for problem solving in practice is therefore largely a matter of experience and it is not realistic to expect a simple look-up table with matches between problems and methods. However, some guidelines can be given and a collection that summarizes other people's experience can also be helpful. A small number of definitions characterize the tasks that are performed by a large proportion of methods. Most of the variation in methods is concerned with differences in data types and algorithmic aspects of methods. In this paper, we summarize the main task types and illustrate how a wide variety of practical problems are formulated in terms of these tasks. The match between problems and tasks is illustrated with a collection of example applications with the aim of helping to express new practical problems as Machine Learning tasks. Some tasks can be decomposed into subtasks, allowing a wider variety of matches between practical problems and (combinations of) methods. We review the main principles for choosing between alternatives and illustrate this with a large collection of applications. We believe that this provides some guidelines.

Ontology has been recently proposed as an appropriate formalism to model software project management antipatterns, in order to encode antipatterns in a computer understandable form and introduce antipatterns to the Semantic Web. However, given two antipattern ontologies, the same entity can be described using different terminology. Therefore, the detection of similar antipattern ontologies is a difficult task. In this paper, we introduce a three-layered antipattern semantic social network, which involves the social network, the antipattern ontology network and the concept network. Social Network Analysis (SNA) techniques can be used to assist software project managers in finding similar antipattern ontologies. For this purpose, SNA measures are extracted from one layer of the semantic social network to another and this knowledge is used to infer new links between antipattern ontologies. The level of uncertainty associated with each new link is represented using Bayesian Networks (BNs). Furthermore, BNs address the issue of quantifying the uncertainty of the data collected regarding antipattern ontologies for the purposes of the conducted analysis. Finally, BNs are used to augment SNA by taking into account meta-information in their calculations. Hence, other knowledge not included in the social network can be used in order to search the social network for further inference. The benefits of using an antipattern semantic social network are illustrated using an example community of software project management antipattern ontologies.

As the number and size of very large databases continues to grow rapidly, so does the need to make sense of them. This need is addressed by the field called knowledge Discovery in Databases (KDD), which combines approaches from machine learning, statistics, intelligent databases, and knowledge acquisition. KDD encompasses a number of different discovery methods, such as clustering, data summarization, learning classification rules, finding dependency networks, analysing changes, and detecting anomalies (Matheus et at., 1993).

The retrieval of relevant cases plays a crucial role in case-based reasoning. There are three major methods for the retrieval of relevant cases: computational approaches (based upon measures of similarity), representational approaches (based upon indexing structures) and hybrid approaches. This paper looks at recent successful implementations of case retrieval with regard to this classification framework. Similarly it emphasises computational and representational models applied to feature-vector case representations.

Constraint satisfaction is a fundamental artificial intelligence technique offering a simple yet powerful representation. An increasing amount of attention has recently been paid to the development of constraint satisfaction techniques, and it has become clear that the original formulation of a static Constraint Satisfaction Problem (CSP) with hard, imperative constraints is insufficient to model many real problems. Two important extensions to the classical CSP framework which address some of these deficiencies are flexible and dynamic constraint satisfaction. This paper examines in detail classical, flexible and dynamic CSP. It reviews the motivations behind both extensions, and describes the techniques used to solve each type of problem. The paper employs a running example throughout to illustrate the ideas presented.

The fast increase in the size and number of databases demands data mining approaches that are scalable to large amounts of data. This has led to the exploration of parallel computing technologies in order to perform data mining tasks concurrently using several processors. Parallelization seems to be a natural and cost-effective way to scale up data mining technologies. One of the most important of these data mining technologies is the classification of newly recorded data. This paper surveys advances in parallelization in the field of classification rule induction.

Artificial neural networks (ANNs) can be employed as controllers for robotic agents. Their structure is often complex, with many neurons and connections, especially when the robots have many sensors and actuators distributed across their bodies and/or when high expressive power is desirable. Pruning (removing neurons or connections) reduces the complexity of the ANN, thus increasing its energy efficiency, and has been reported to improve the generalization capability, in some cases. In addition, it is well-known that pruning in biological neural networks plays a fundamental role in the development of brains and their ability to learn. In this study, we consider the evolutionary optimization of neural controllers for the case study of Voxel-based soft robots, a kind of modular, bio-inspired soft robots, applying pruning during fitness evaluation. For a locomotion task, and for centralized as well as distributed controllers, we experimentally characterize the effect of different forms of pruning on after-pruning effectiveness, life-long effectiveness, adaptability to new terrains, and behavior. We find that incorporating some forms of pruning in neuroevolution leads to almost equally effective controllers as those evolved without pruning, with the benefit of higher robustness to pruning. We also observe occasional improvements in generalization ability.

Academic work on agents and ontologies is often oblivious to the complexities and realities of enterprise computing. At the same time, the practitioners of enterprise computing, although they are adept at the building of robust, real-life enterprise applications, are unaware of the academic body of work and the opportunities of applying novel approaches of academic origin. Enterprise applications are very complex systems that are designed to support critical business operations. This article outlines the technical and business foundations of enterprise application software and briefly discusses viable opportunities for agents and ontology research.

The promise of rule-based computing was to allow end-users to create, modify, and maintain applications without the need to engage programmers. But experience has shown that rule sets often interact in subtle ways, making them difficult to understand and reason about. This has impeded the widespread adoption of rule-based computing. This paper describes the design and implementation of XcelLog, a user-centered deductive spreadsheet system, to empower non-programmers to specify and manipulate rule-based systems. The driving idea underlying the system is to treat sets as the fundamental data type and rules as specifying relationships among sets, and use the spreadsheet metaphor to create and view the materialized sets. The fundamental feature that makes XcelLog suitable for non-programmers is that the user mainly sees the effect of the rules; when rules or basic facts change, the user sees the impact of the change immediately. This enables the user to gain confidence in the rules and their modification, and also experiment with what-if scenarios without any programming. Preliminary experience with using XcelLog indicates that it is indeed feasible to put the power of deductive spreadsheets for doing rule-based computing into the hands of end-users and do so without the requirement of programming or the constraints of canned application packages.