.Introduction to the papers presented at the 2013 FLAIRS Conference Special Track on Applied Natural Language Processing.

A conditional CSP is a well known formalism for managing constraints in a dynamic environment where the possible changes are known a priori and can be enumerated beforehand. In this paper we argue that mixing preferences with Conditional CSPs brings more intuitive meaning to many real world applications. For example, in a product configuration problem the user preferences often co-exist with the configuration requirements in a dynamic environment. We propose here an extension of the Conditional CSP formalism in order to consider qualitative preferences. We use CP-nets as a representation for these preferences and define new concepts for preserving their semantics in dynamic settings.

In this study, we propose a model which may assist in diagnosis of Alzheimer’s disease (AD) using T1 weighted MRI brain images. The proposed model involves construction of statistical features from multiple trans-axial slices from hippocampus and amygdala regions, which play a significant role in AD diagnosis. Features from multiple slices are then averaged, which resulted into a smaller set of relevant features. The reduced set of features enhances the performance of decision learning system, and takes less memory and computation time. Effectiveness of the proposed model is compared with recent voxel-based-morphometry work in terms of sensitivity, specificity and accuracy. Experimental results on a publicly available MRI dataset showed that the proposed method outperforms the recent voxel-based-morphometry model.

High-dimensional indexing is an important area of current research, especially for range and kNN queries. This work introduces clustering for the sake of indexing. The goal is to develop new clustering methods designed to optimize the data partitioning for an indexing-specific tree structure instead of finding data distribution-based clusters. We focus on iDistance, a state-of-the-art high-dimensional indexing method, and take a basic approach to solving this new problem. By utilizing spherical clusters in an unsupervised Expectation Maximization algorithm dependent upon local density and cluster overlap, we create a partitioning of the space providing balanced segmentation for a B+-tree. We also look at the novel idea of reclustering for a specific indexing method by taking the output of one clustering method and reclustering it for use in an index. The algorithms are then tested and evaluated based on our error metric and iDistance query performance.

Goal-driven autonomy is a framework for intelligent agents that automatically formulate and manage goals in dynamic environments, where goal formulation is the task of identifying goals that the agent should attempt to achieve. We argue that goal formulation is central to high-level autonomy, and explain why identifying domain-independent heuristics for this task is an important research topic in high-level control. We describe two novel domain-independent heuristics for goal formulation (motivators) that evaluate the utility of goals based on the projected consequences of achieving them. We then describe their integration in M-ARTUE, an agent that balances the satisfaction of internal needs with the achievement of goals introduced externally. We assess its performance in a series of experiments in the Rovers With Compass domain. Our results show that using domain-independent heuristics yields performance comparable to using domain-specific knowledge for goal formulation. Finally, in ablation studies we demonstrate that each motivator contributes significantly to M-ARTUE’s performance.

While researching spatial movements in play-scripts, we uncovered some movements that actors performed that could not be explained by the annotations or basic theatre rules. Here, we look to learn implied motion based on what the characters in the play are saying. Humans are able to do this with only being given the play-script, so how do we get a computer to do it? Several features, including n -grams, parts of speech (POS) bag of words (BoW), length of speech, and other contextual details were utilized with several machine learning methods to help predict movement within the play. Results reveal that is a difficult problem and basic natural language processing (NLP) and machine learning (ML) techniques do not perform much better than a random classifier.

We propose and develop a simple and efficient algorithm for generating extractive multi-document summaries and show that this algorithm exhibits state-of-the-art or near state-of-the-art performance on two Document Understanding Conference datasets and two Text Analysis Conference datasets. Our results show that algorithms using simple features and computationally efficient methods are competitive with much more complex methods for multi-document summarization (MDS). Given these findings, we believe that our summarization algorithm can be used as a baseline in future MDS evaluations. Further, evidence shows that our system is near the upper limit of performance for extractive MDS.

Emotion regulation looks into methods and strategiesthat humans use in order to control and balance theirpossible extreme levels of emotions. One importantchallenge in building a computational model of emotionsis the mainly non-quantitative nature of this problem.In this paper, we investigate a Fuzzy logic approachas a possible framework for providing the requiredqualitative and quantitative description of suchmodels. In our proposed fuzzy computational modelwhich was constructed based on Gross theory for emotionregulation, beside the fuzzy structure, it includesa learning module that enhances the model adaptivityto environmental changes through learning some relevantaspects such as patterns of events’ sequences. Theresults of the simulation experiments were comparedagainst a formerly presented non-fuzzy implementation.We observed that the agents in our proposed modelmanaged to cope better with changes in the environmentand exhibited smoother regulation behavior. Moreover,our model showed further consistency with the inferentialrules of Gross theory.

Support vector machines (SVMs) are among the most used methods for pattern recognition. Acceptable results have been obtained with such methods in many domains and applications. However, as most learning algorithms, SVMs have hyperparameters that influence the effectiveness of the generated model. Thus, choosing adequate values for such hyperparameters is critical in order to obtain satisfactory results for a given classification task, a problem known as model selection. This paper introduces a novel model selection approach for SVMs based on multi-objective optimization and on the bias and variance definition. We propose an evolutionary algorithm that aims to select the configuration of hyperparameters that optimizes a trade-off between estimates of bias and variance; two factors that are closely related to the model accuracy and complexity. The proposed technique is evaluated using a suite of benchmark data sets for classification. Experimental results show the validity of our approach. We found that the model selection criteria resulted very helpful for selecting highly effective classification models.

In this paper, we develop and demonstrate a new 2nd order two-stage algorithm called OWO-Newton. First, two-stage algorithms are motivated and the Gauss Newton input weight Hessian matrix is developed. Block coordinate descent is used to apply Newton’s algorithm alternately to the input and output weights. Its performance is comparable to Levenberg-Marquardt and it has the advantage of reduced computational complexity. The algorithm is shown to have a form of affine invariance.