Progress in Artificial Intelligence

E-Learning, much like any other communication processes, has been significantly shaped by technological evolution. In its original form, e-Learning aimed to bring the education closer to people, making it more modular and personalized. However, in reality, we observe that it represents a separation between student and teacher, simplifying this relationship to the exchange of ”text-based messages”, leaving aside all the important contextual richness of the classroom. We are addressing this issue by devising a contextual layer for e-Learning platforms. Particularly, in this paper we describe a solution to convey information about the level of stress of the students so that the teacher can take better and more informed decisions concerning the management of the learning process.

Nowadays, a number of computing paradigms have been proposed, of which the latest one is known as Cloud computing.Cloud computing is revolutionizing the services provided through the Internet, and is continually adapting itself in order to maintain the quality of its services. In this paper is proposes a cloud platform for storing information and files by following the cloud paradigm. Moreover, a cloud-based application has been developed to validate the services provided by the platform.

Artificial Neural Networks (ANN) have been essentially used as regression models to predict the concentration of one or more pollutants usually requiring information collected from air quality stations. In this work we consider a Multilayer Perceptron (MLP) with one hidden layer as a classifier of the impact of air quality on human health, using only traffic and meteorological data as inputs. Our data was obtained from a specific urban area and constitutes a 2-class problem: above or below the legal limits of specific pollutant concentrations. The results show that an MLP with 40 to 50 hidden neurons and trained with the cross-entropy cost function, is able to achieve a mean error around 11%, meaning that air quality impacts can be predicted with good accuracy using only traffic and meteorological data. The use of an ANN without air quality inputs constitutes a significant achievement because governments may therefore minimize the use of such expensive stations.

Difficulty faced by drivers in finding a parking space in either car parks or in the street is one of the common problems shared by all the big cities, most of the times leading to traffic congestion and driver frustration. Exploiting the capabilities that Computer Vision offers, an alternative to those ITS commercial solutions for parking space detection that rely on other sensors different from cameras is presented. The system is able to detect vacant spaces and classify them by the type of vehicle that could park in that area. First of all, an approximate inverse perspective transformation is applied for 2D to 3D reconstruction of parking area. In addition, feature analysis based on Pyramid Histogram of Oriented Gradients (PHOG) is carried out on every parking zone within the parking area. Experiments on real scenarios show the excellent capabilities of the proposed system with independence of camera orientation in the context.

In evolutionary game theory, one is normally interested in the investigation about how the distribution of strategies changes along time. Equilibrium-based methods are not appropriate for open, dynamic systems, as for instance those in which individual drivers learn to select routes. In this paper we model route choice in which many agents adapt simultaneously. We investigate the dynamics with a continuous method (replicator dynamics), and with learning methods (social and individual). We show how the convergence to one of the Nash equilibria depends on the underlying learning dynamics selected, as well as on the pace of adjustments by the driver agents.

Informed driving

is becoming a key feature to increase the sustainability of taxi companies. Some recent works are exploring the data broadcasted by each vehicle to provide live information for decision making. In this paper, we propose a method to employ a learning model based on historical GPS data in a real-time environment. Our goal is to predict the spatiotemporal distribution of the Taxi-Passenger demand in a short time horizon. We did so by using learning concepts originally proposed to a well-known online algorithm: the

perceptron

[1]. The results were promising: we accomplished a satisfactory performance to output the next prediction using a short amount of resources.

In this paper a novel approach to a decentralized autonomous model of agency for general purpose Non-Player Characters (NPCs) is presented: the AI model of Computational Ecosystems. We describe the technology used to animate a population of gregarious humanoid avatars in a virtual world. This artistic work is an ethnographic project where a population of NPCs inhabit the virtual world and interact autonomously among themselves as well as with an audience of outsiders (human observers). First, we present the background, motivation and summary for the project. Then, we describe the algorithm that was developed to generate the movements and behaviors of the population of NPC “story-tellers”. Finally, we discuss some of the critical aspects of this implementation and contextualize the work with regards to a wider usage in computer games and virtual worlds.

The purpose of this study is to develop an innovative system for Coumarin-derived drug dosing, suitable for elderly patients. Recent research highlights that the pharmacological response of the patient is often affected by many exogenous factors other than the dosage prescribed and these factors could form a very complex relationship with the drug dosage. For this reason, new powerful computational tools are needed for approaching this problem. The system we propose is called Geometric Semantic Genetic Programming, and it is based on the use of recently defined geometric semantic genetic operators. In this paper, we present a new implementation of this Genetic Programming system, that allow us to use it for real-life applications in an efficient way, something that was impossible using the original definition. Experimental results show the suitability of the proposed system for managing anticoagulation therapy. In particular, results obtained with Geometric Semantic Genetic Programming are significantly better than the ones produced by standard Genetic Programming both on training and on out-of-sample test data.

In this paper, we investigate the dynamics of different neuronal models on online neuroevolution of robotic controllers in multirobot systems. We compare the performance and robustness of neural network-based controllers using summing neurons, multiplicative neurons, and a combination of the two. We perform a series of simulation-based experiments in which a group of e-puck-like robots must perform an integrated navigation and obstacle avoidance task in environments of different complexity. We show that: (i) multiplicative controllers and hybrid controllers maintain stable performance levels across tasks of different complexity, (ii) summing controllers evolve diverse behaviours that vary qualitatively during task execution, and (iii) multiplicative controllers lead to less diverse and more static behaviours that are maintained despite environmental changes. Complementary, hybrid controllers exhibit both behavioural characteristics, and display superior generalisation capabilities in simple and complex tasks.

Evolutionary Algorithms (EA) are loosely inspired in the ideas of natural selection and genetics. Over the years some researchers have advocated the need of incorporating more ideas from biology into EAs, in particular with respect to the individuals’ representation and the mapping from the genotype to the phenotype. One of the first successful proposals in that direction was the Artificial Regulatory Network (ARN) model. Soon after some variants of the ARN with increased capabilities were proposed, namely the Regulatory Network Computational Device (ReNCoDe). In this paper we further study ReNCoDe, testing the implications of some design choices of the underlying ARN model. A Genetic Programming community-approved symbolic regression benchmark (the harmonic number) is used to compare the performance of the different alternatives.

The performance of an evolutionary algorithm strongly depends on the choice of the parameters which regulate its behavior. In this paper, two evolutionary algorithms (Particle Swarm Optimization and Differential Evolution) are used to find the optimal configuration of parameters for Differential Evolution. We tested our approach on four benchmark functions, and the comparison with an exhaustive search demonstrated its effectiveness. Then, the same method was used to tune the parameters of Differential Evolution in solving a real-world problem: the automatic localization of the hippocampus in histological brain images. The results obtained consistently outperformed the ones achieved using manually-tuned parameters. Thanks to a GPU-based implementation, our tuner is up to 8 times faster than the corresponding sequential version.

Evolving constants in Genetic Programming is still an open issue. As real values they cannot be integrated in GP trees in a direct manner, because the nodes represent discrete symbols. Present solutions are the concept of ephemeral random constants or hybrid approaches, which have additional computational costs. Furthermore, one has to change the GP algorithm for them. This paper proposes a concept, which does not change the GP algorithm or its components. Instead, it introduces structure-based constants realized as functions, which can be simply added to each function set while keeping the original GP approach. These constant functions derive their constant values from the tree structures of their child-trees (subtrees). That is, a constant is represented by a tree structure being this way under the influence of the typical genetic operators like subtree crossover or mutation. These structure-based constants were applied to symbolic regression problems. They outperformed the standard approach of ephemeral random constants. Their results together with their better properties make the structure-based constant concept a possible candidate for the replacement of the ephemeral random constants.

Class imbalance affects medical diagnosis, as the number of disease cases is often outnumbered. When it is severe, learning algorithms fail to retrieve the rarer classes and common assessment metrics become uninformative. In this work, class imbalance is approached using neuropsychological data, with the aim of differentiating Alzheimer’s Disease (AD) from Mild Cognitive Impairment (MCI) and predicting the conversion from MCI to AD. The effect of the imbalance on four learning algorithms is examined through the application of bagging, Bayes risk minimization and MetaCost. Plain decision trees were always outperformed, indicating susceptibility to the imbalance. The naïve Bayes classifier was robust but suffered a bias that was adjusted through risk minimization. This strategy outperformed all other combinations of classifiers and meta-learning/ensemble methods. The tree-augmented naïve Bayes classifier also benefited from an adjustment of the decision threshold. In the nearly balanced datasets, it was improved by bagging, suggesting that the tree structure was too strong for the attribute dependencies. Support vector machines were robust, as their plain version achieved good results and was never outperformed.

High-throughput technologies such as microarray data are a great resource for studying and understanding biological systems at a low cost. However noise present in the data makes it less reliable, and thus many computational methods and algorithms have been developed for removing the noise. We propose a novel noise removal algorithm based on Fourier transform functions. The algorithm optimizes the coefficients of the first and second order Fourier functions and selects the function which maximizes the Spearman correlation to the original data. To demonstrate the performance of this algorithm we compare the prediction accuracy of well known modelling tools, such as network component analysis (NCA), principal component analysis (PCA) and k-means clustering. We compared the performance of these tools on the original noisy data and the data treated with the algorithm. We performed the comparison analysis using three independent real biological data sets (each data set with two replicates). In all cases the proposed algorithm removes the noise in the data and substantially improves the predictions of modelling tools.

In this paper we propose a novel

associative metric

based on the classical conditioning paradigm that, much like what happens in nature, identifies associations between stimuli perceived by a learning agent while interacting with the environment. We use an associative tree structure to identify associations between the perceived stimuli and use this structure to measure the degree of similarity between states in factored Markov decision problems. Our approach provides a

state-space metric

that requires no prior knowledge on the structure of the underlying decision problem and is designed to be learned online,

i.e.

, as the agent interacts with its environment. Our metric is thus amenable to application in reinforcement learning (RL) settings, allowing the learning agent to generalize its experience to unvisited states and improving the overall learning performance. We illustrate the application of our method in several problems of varying complexity and show that our metric leads to a performance comparable to that obtained with other well-studied metrics that require full knowledge of the decision problem.

Generating diagnosis candidates for a set of failing transactions is an important challenge in the context of automatic fault localization of both software and hardware systems. Being an NP-Hard problem, exhaustive algorithms are usually prohibitive for real-world, often large, problems. In practice, the usage of heuristic-based approaches trade-off completeness for time efficiency. An example of such heuristic approaches is

Staccato

, which was proposed in the context of reasoning-based fault localization. In this paper, we propose an efficient distributed algorithm, dubbed MHS

2

, that renders the sequential search algorithm

Staccato

suitable to distributed, Map-Reduce environments. The results show that MHS

2

scales to larger systems (when compared to

Staccato

), while entailing either marginal or small runtime overhead.

In this paper, a hybrid clustering and classification algorithm is obtained by exploring the specific statistical model of a hyperplane classifier. We show how the seamless integration of the clustering component allows a substantial cost decrease in the training stage, without impairing the performance of the classifier. The algorithm is also robust to outliers and deals with training errors in a natural and efficient manner.

This paper presents an approach, based on propositional satisfiability (SAT), for the resource constrained project scheduling problem with general precedence relations. This approach combines propositional satisfiability formulations with a bisection method, in order to achieve an optimal solution. The empirical results suggest that when the optimal schedule is significantly affected by the availability of resources, this strategy outperforms the typical integer linear programming approach.

A binary classification algorithm, called Probabilistic Vector Machine – PVM, is proposed. It is based on statistical measurements of the training data, providing a robust and lightweight classification model with reliable performance. The proposed model is also shown to provide the optimal binary classifier, in terms of probability of error, under a set of loose conditions regarding the data distribution. We compare PVM against GEPSVM and PSVM and provide evidence of superior performance on a number of datasets in terms of average accuracy and standard deviation of accuracy.

Despite its potential as a reasoning paradigm in AI applications, abduction has been on the back burner in logic programming, as abduction can be too difficult to implement, and costly to perform, in particular if abductive solutions are not tabled. If they become tabled, then abductive solutions can be reused, even from one abductive context to another. On the other hand, current Prolog systems, with their tabling mechanisms, are mature enough to facilitate the introduction of tabling abductive solutions (tabled abduction) into them. The concept of tabled abduction has been realized recently in an abductive logic programming system

tabdual

. Besides tabling abductive solutions,

tabdual

also relies on the dual transformation. In this paper, we emphasize two

tabdual

improvements: (1) the dual transformation

by need

, and (2) a new construct for accessing ongoing abductive solutions, that permits modular mixes between abductive and non-abductive program parts. We apply subsequently these improvements on two distinct problems, and evaluate the performance and the scalability of

tabdual

on several benchmarks on the basis of these problems, by examining four

tabdual

variants.

The detection of the ball when it is not on the ground is an important research line within the Middle Size League of RoboCup. A correct detection of airborne balls is particularly important for goal keepers, since shots to goal are usually made that way. To tackle this problem on the CAMBADA team , we installed a perspective camera on the robot. This paper presents an analysis of the scenario and assumptions about the use of a single perspective camera for the purpose of 3D ball perception. The algorithm is based on physical properties of the perspective vision system and an heuristic that relates the size and position of the ball detected in the image and its position in the space relative to the camera. Regarding the ball detection, we attempt an approach based on a hybrid process of color segmentation to select regions of interest and statistical analysis of a global shape context histogram. This analysis attempts to classify the candidates as round or not round. Preliminary results are presented regarding the ball detection approach that confirms its effectiveness in uncontrolled environments. Moreover, experimental results are also presented for the ball position estimation and a sensor fusion proposal is described to merge the information of the ball into the worldstate of the robot.

In this paper we introduce a novel image descriptor, LBP-gist, suitable for real time loop closure detection. As the name suggests, the proposed method builds on two popular image analysis techniques: the gist feature, which has been used in holistic scene description and the LBP operator, originally designed for texture classification. The combination of the two methods gives rise to a very fast computing feature which is shown to be competitive to the state-of-the-art loop closure detection. Fast image search is achieved via Winner Take All Hashing, a simple method for image retrieval that exploits the descriptive power of rank-correlation measures. Two modifications of this method are proposed, to improve its selectivity. The performance of LBP-gist and the hashing strategy is demonstrated on two outdoor datasets.

Most of the original image feature detectors are not able to cope with large photometric variations, and their extensions that should improve detection eventually increase the computational cost and introduce more noise to the system. Here we extend the original SURF algorithm increasing its invariance to illumination changes. Our approach uses the local space average color descriptor as working space to detect invariant features. A theoretical analysis demonstrates the impact of distinct photometric variations on the response of blob-like features detected with the SURF algorithm. Experimental results demonstrate the effectiveness of the approach in several illumination conditions including the presence of two or more distinct light sources, variations in color, in offset and scale.

Assistive Technologies may greatly contribute to give autonomy and independence for individuals with physical limitations. Electric wheelchairs are examples of those assistive technologies and nowadays each time becoming more intelligent due to the use of technology that provides assisted safer driving. Usually, the user controls the electric wheelchair with a conventional analog joystick. However, this implies the need for an appropriate methodology to map the position of the joystick handle, in a Cartesian coordinate system, to the wheelchair wheels intended velocities. This mapping is very important since it will determine the response behavior of the wheelchair to the user manual control. This paper describes the implementation of several joystick mappings in an intelligent wheelchair (IW) prototype. Experiments were performed in a realistic simulator using cerebral palsy users with distinct driving abilities. The users had 6 different joystick control mapping methods and for each user the usability and the users’ preference order was measured. The results achieved show that a linear mapping, with appropriate parameters, between the joystick’s coordinates and the wheelchair wheel speeds is preferred by the majority of the users.

Soccer Humanoid robots must be able to fulfill their tasks in a highly dynamic soccer field, which requires highly responsive and dynamic locomotion. It is very difficult to keep humanoids balance during walking. The position of the Zero Moment Point (ZMP) is widely used for dynamic stability measurement in biped locomotion. In this paper, we present an omnidirectional walk engine, which mainly consist of a Foot planner, a ZMP and Center of Mass (CoM) generator and an Active balance loop. The Foot planner, based on desire walk speed vector, generates future feet step positions that are then inputs to the ZMP generator. The cart-table model and preview controller are used to generate the CoM reference trajectory from the predefined ZMP trajectory. An active balance method is presented which keeps the robot’s trunk upright when faced with environmental disturbances. We have tested the biped locomotion control approach on a simulated NAO robot. Our results are encouraging given that the robot has been able to walk fast and stably in any direction with performances that compare well to the best RoboCup 2012 3D Simulation teams.

This paper presents a scan-matching approach for online simultaneous localization and mapping. This approach combines a fast and efficient scan-matching algorithm for localization with dynamic and approximate likelihood fields to incrementally build a map. The achievable results of the approach are evaluated using an objective benchmark designed to compare SLAM solutions that use different methods. The result is a fast online SLAM approach suitable for real-time operations.

Cooperation with humans is a requirement for the next generation of robots so it is necessary to model how robots can sense, know, share and acquire knowledge from human interaction. Instead of traditional SLAM (Simultaneous Localization and Mapping) methods, which do not interpret sensor information other than at the geometric level, these capabilities require an environment map representation similar to the human representation. Topological maps are one option to translate these geometric maps into a more abstract representation of the the world and to make the robot knowledge closer to the human perception. In this paper is presented a novel approach to translate 3D grid map into a topological map. This approach was optimized to obtain similar results to those obtained when the task is performed by a human. Also, a novel feature of this approach is the augmentation of topological map with features such as walls and doors.

A formation driving mechanism suited for utilization of multi-robot teams in highly dynamic environments is proposed in this paper. The presented approach enables to integrate a prediction of behaviour of moving objects in robots’ workspace into a formation stabilization and navigation framework. It will be shown that such an inclusion of a model of the surrounding environment directly into the formation control mechanisms facilitates avoidance manoeuvres in a case of fast dynamic objects approaching in a collision course. Besides, the proposed model predictive control based approach enables to stabilize robots in a compact formation and it provides a failure tolerance mechanism with an inter collision avoidance. The abilities of the algorithm are verified via numerous simulations and hardware experiments with the main focus on evaluation of performance of the algorithm with different sensing capabilities of the robotic system.

In data clustering, the problem of selecting the subset of most relevant features from the data has been an active research topic. Feature selection for clustering is a challenging task due to the absence of class labels for guiding the search for relevant features. Most methods proposed for this goal are focused on numerical data. In this work, we propose an approach for clustering and selecting categorical features simultaneously. We assume that the data originate from a finite mixture of multinomial distributions and implement an integrated expectation-maximization (EM) algorithm that estimates all the parameters of the model and selects the subset of relevant features simultaneously. The results obtained on synthetic data illustrate the performance of the proposed approach. An application to real data, referred to official statistics, shows its usefulness.

Monitoring the habits of elderly people is a great challenge in order to improve ageing at home. Studying the deviances from or the evolution of regular behaviors may help to detect emerging pathologies. Regular patterns are searched in the data coming from sensors disseminated in the elderly’s home. An efficient algorithm, xED, is proposed to mine such patterns. It emphasizes the description of the variability in the times when habits usually occur, and is robust to parasite events. Experiment on real-life data shows the interest of xED.

In this study, we explored data from StockTwits, a microblogging platform exclusively dedicated to the stock market. We produced several indicators and analyzed their value when predicting three market variables: returns, volatility and trading volume. For six major stocks, we measured posting volume and sentiment indicators. We advance on the previous studies on this subject by considering a large time period, using a robust forecasting exercise and performing a statistical test of forecasting ability. In contrast with previous studies, we find no evidence of return predictability using sentiment indicators, and of information content of posting volume for forecasting volatility. However, there is evidence that posting volume can improve the forecasts of trading volume, which is useful for measuring stock liquidity (e.g. assets easily sold).

Predicting the future impact of academic publications has many important applications. In this paper, we propose methods for predicting future article impact, leveraging digital libraries of academic publications containing citation information. Using a set of successive past impact scores, obtained through graph-ranking algorithms such as PageRank, we study the evolution of the publications in terms of their yearly impact scores, learning regression models to predict the future PageRank scores, or to predict the future number of downloads. Results obtained over a DBLP citation dataset, covering papers published up to the year of 2011, show that the impact predictions are highly accurate for all experimental setups. A model based on regression trees, using features relative to PageRank scores, PageRank change rates, author PageRank scores, and term occurrence frequencies in the abstracts and titles of the publications, computed over citation graphs from the three previous years, obtained the best results.

Several real world prediction problems involve forecasting rare values of a target variable. When this variable is nominal we have a problem of class imbalance that was already studied thoroughly within machine learning. For regression tasks, where the target variable is continuous, few works exist addressing this type of problem. Still, important application areas involve forecasting rare extreme values of a continuous target variable. This paper describes a contribution to this type of tasks. Namely, we propose to address such tasks by sampling approaches. These approaches change the distribution of the given training data set to decrease the problem of imbalance between the rare target cases and the most frequent ones. We present a modification of the well-known

Smote

algorithm that allows its use on these regression tasks. In an extensive set of experiments we provide empirical evidence for the superiority of our proposals for these particular regression tasks. The proposed

SmoteR

method can be used with any existing regression algorithm turning it into a general tool for addressing problems of forecasting rare extreme values of a continuous target variable.

The analysis of frequent behaviors regarding temporal issues begins to achieve some interest, in particular in the area of health care. However, existing approaches tend to ignore the temporal information and only make use of the order among events occurrence. In this paper, we introduce the notion of temporal constraint, and propose three instantiations of it: complete cyclic temporal constraints, partial cyclic temporal constraints and timespan constraints. Additionally, we propose a new algorithm – (TD)

2

PaM, that together with these constraints, makes possible to focus the pattern mining process on looking for cyclic and timespan patterns. Experimental results reveal the algorithm to be as efficient as its predecessors, and able to discover more informed patterns.

One research challenge in multi-agent systems is how to partition a set of agent into coalition structures. The number of different coalitions for a set of agents grows exponential with the number of agents and the number of different partitions grows even faster. The common approach for this problem is to search for the coalition structure that maximizes the system outcome (denoted

CS

*

). Until recently, most algorithms that solve this problem considered that the value of a coalition is given by a characteristic function, in where those values are not influenced by factors that are external to the coalition. More recently, several authors have focused on problems that consider the presence of externalities. In this case, centralized algorithms were developed to search for the

CS

*

, but no algorithm was developed to work in a distributed environment. This paper presents a distributed algorithm for searching coalition structures under presence of externalities.

Ambient Intelligence (AmI) is the vision of a future in which environments support the people inhabiting them unobtrusively. The distributed nature the AmI, the autonomy, awareness and adaptation properties make software agents a good option to develop self-managed AmI systems. AmI systems demand the reconfiguration of their internal functioning in response to changes in the environment. So, AmI systems must behave as autonomic systems with the capacity for self-management. However, current agent metamodels lack adequate modeling mechanisms to cope with self-management. We propose to extend an existing agent metamodel with a number of specific modeling concepts and elements that facilitate the design of the self-management capabilities in agents.

This paper introduces an operational semantics for defining Intentional Learning on

Jason

, the well known Java-based implementation of

AgentSpeak

(

L

). This semantics enables

Jason

to define agents capable of learning the reasons for adopting intentions based on their own experience. In this work, the use of the term

Intentional Learning

is strictly circumscribed to the practical rationality theory where plans are predefined and the target of the learning processes is to learn the reasons to adopt them as intentions. Top-Down Induction of Logical Decision Trees (TILDE) has proved to be a suitable mechanism for supporting learning on

Jason

: the first-order representation of TILDE is adequate to form training examples as sets of beliefs, while the obtained hypothesis is useful for updating the plans of the agents.

In this paper we present a SAT-based Bounded Model Checking (BMC) method for weighted deontic interpreted systems (i.e., Kripke structures where transitions carry a weight, which is an arbitrary natural number) and properties expressed in the existential fragment of a weighted temporal logic augmented to include knowledge and deontic components (

Wectlkd

). In particular, since in BMC both the system model and the checked property are translated into a Boolean formula to be analysed by a SAT-solver, we introduce a new Boolean encoding of the

Wectlkd

formulae that is particularly optimized for managing quantitative weighted temporal operators, knowledge operators, and deontic operators, which are typically found in properties of complex multi-agent systems in models of which we assume the possibility that agents may not behave as they are supposed to, and that acting (coordination, negotiation, cooperation, etc.) of agents may cost. We illustrate how the weighted deontic interpreted systems can be applied to the analysis of a variant of the standard bit transmission problem in which an agent may fail to do something it is supposed to do.

In real world scenarios, the formation of consensus is an self-organisation process by which actors have to make a joint assessment about a target subject being it a decision making problem or the formation of a collective opinion. In social simulation, models of opinion dynamics tackle the opinion formation phenomena. These models try to make an assessment, for instance, of the ideal conditions that lead an interacting group of agents to opinion consensus, polarisation or fragmentation. In this paper, we investigate the role of social relation structure in opinion dynamics using an interaction model of relative agreement. We present an agent-based model that defines social relations as multiple concomitant social networks and apply our model to an opinion dynamics model with bounded confidence. We discuss the influence of complex social network topologies where actors interact in multiple relations simultaneously. The paper builds on previous work about social space design with multiple contexts and context switching, to determine the influence of such complex social structures in a process such as opinion formation.

Intelligent systems should be able to initiate and understand social interactions, which are always brought about by physical actions. Thus, all physical actions need to be interpreted for their social effects. This paper gives an analytical framework that identifies the main challenges in modelling social interactions. Such interpretation is subjective and might lead to misunderstandings between participants in the interaction. In real life, people use rituals, such as greetings, as a method for common grounding. Greeting as an essential prerequisite of joint activity is taken as case study to illustrate our approach.

We present initial efforts made to model the spread of the drywood termite in Angra do Heroísmo, Azores, using an agent based modeling approach. First we describe how a simple Cellular Automata (CA) model was created in Netlogo to simulate the spread of the species based on simple assumptions concerning the ecology of the species. A second step was taken by increasing the complexity of the initial CA approach, adding new specific characteristics to each cell, based again on ecology of the species and its behavior towards the environment. Finally, we add agents to the model in order to simulate the human intervention in fighting back the pest. This new model has become a two-level Agent-Based model. We also evaluated the costs of this intervention. These efforts were supported by field research which allowed a continuous cross-checking of the results obtained in the model with the field data.

Keywords are single and multiword terms that describe the semantic content of documents. They are useful in many applications, such as document searching and indexing, or to be read by humans. Keywords can be explicit, by occurring in documents, or implicit, since, although not explicitly written in documents, they are semantically related to their contents. This paper presents a statistical approach to build document descriptors with

explicit

and

implicit

keywords automatically extracted from the documents. Our approach is language-independent and we show comparative results for three different European languages.

Translation tasks, including bilingual concordancing, demand an efficient space/time trade-off, which is not always easy to get due to the usage of huge text collections and the space consuming nature of time efficient text indexes. We propose a compact representation for monotonically aligned parallel texts, based on known compressed text indexes for representing the texts and additional uncompressed structures for the alignment. The proposed framework is able to index a collection of texts in main memory, occupying less space than the text size and with efficient query response time. The proposal supports any type of alignment granularity, a novelty in concordancing applications, allowing a flexible environment for linguistics working in all phases of a translation process. We present two alternatives for self-indexing the texts, and two alternatives for supporting the alignment, comparing the alternatives in terms of space/time performance.

The extraction of semantic relations from texts is currently gaining increasing interest. However, a large number of current methods are language and domain dependent, and the statistical and language-independent methods tend to work only with large amounts of text. This leaves out the extraction of semantic relations from standalone documents, such as single documents of unique subjects, reports from very specific domains, or small books.

We propose a statistical method to extract semantic relations using clusters of concepts. Clusters are areas in the documents where concepts occur more frequently. When clusters of different concepts occur in the same areas, they may represent highly related concepts.

Our method is language independent and we show comparative results for three different European languages.

Sentence Reduction has recently received a great attention from the research community of Automatic Text Summarization. Sentence Reduction consists in the elimination of sentence components such as words, part-of-speech tags sequences or chunks without highly deteriorating the information contained in the sentence and its grammatical correctness. In this paper, we present an unsupervised scalable methodology for learning sentence reduction rules. Paraphrases are first discovered within a collection of automatically crawled Web News Stories and then textually aligned in order to extract interchangeable text fragment candidates, in particular

reduction cases

. As only positive examples exist,

Inductive Logic Programming

(ILP) provides an interesting learning paradigm for the extraction of sentence

reduction rules

. As a consequence, reduction cases are transformed into first order logic clauses to supply a massive set of suitable learning instances and an ILP learning environment is defined within the context of the

Aleph

framework. Experiments evidence good results in terms of irrelevancy elimination, syntactical correctness and reduction rate in a real-world environment as opposed to other methodologies proposed so far.

The Time-series microarray data that we used to demonstrate the algorithm was received from the groups of Prof. A. Lægreid and Prof. M. Kuiper. In the original submission we accidently referred to the source of the raw data, GSE32869 and GSE13009, but we actually received and used modified datasets. Specifically, in the Methods section we write, ≪Prior to the expression data fitting, we performed the selection of the differentially expressed genes based on fold change and p-value including normalization.≫, whereas this selection actually was performed by other groups at our University. We would like therefore to add the following acknowledgements:

The authors N.D.J and N.B. would like to thank Prof. A. Laegreid, NTNU, Prof. M. Kuiper, NTNU, their research groups, and their associates at the St. Olav’s hospital, for providing us with the time series microarray data used to demonstrate the noise removal algorithm.