Abstract
The ensembling of classifiers tends to improve predictive accuracy. To obtain an ensemble with N classifiers, one typically needs to run N learning processes. In this paper we introduce and explore Model Jittering Ensembling, where one single model is perturbed in order to obtain variants that can be used as an ensemble. We use as base classifiers sets of classification association rules. The two methods of jittering ensembling we propose are Iterative Reordering Ensembling (IRE) and Post Bagging (PB). Both methods start by learning one rule set over a single run, and then produce multiple rule sets without relearning. Empirical results on 36 data sets are positive and show that both strategies tend to reduce error with respect to the single model association rule classifier. A bias-variance analysis reveals that while both IRE and PB are able to reduce the variance component of the error, IRE is particularly effective in reducing the bias component. We show that Model Jittering Ensembling can represent a very good speed-up w.r.t. multiple model learning ensembling. We also compare Model Jittering with various state of the art classifiers in terms of predictive accuracy and computational efficiency.

Abstract
This paper plans an end-to-end method for extracting information from tables embedded in documents; input format is ASCII, to which any richer fort-nat can be converted, preserving all textual and much of the layout information. We start by defining table. Then we describe the steps involved in extracting information from tables and analyse table-related research to place the contribution of different authors, find the paths research is following, and identify issues that are still unsolved. We then analyse current approaches to evaluating table processing algorithms and propose two new metrics for the task of segmenting cells/columns/rows. We proceed to design our own end-to-end method, where there is a higher interaction between different steps; we indicate how back loops in the usual order of the steps can reduce the possibility of errors and contribute to solving previously unsolved problems. Finally, we explore how the actual interpretation of the table not only allows inferring the accuracy of the overall extraction process but also contributes to actually improving its quality. In order to do so, we believe interpretation has to consider context-specific knowledge; we explore how the addition of this knowledge can be made in a plug-in/out manner, such that the overall method will maintain its operability in different contexts.

Abstract
In this paper we study the deviation of bus trip duration and its causes. Deviations are obtained by comparing scheduled times against actual trip duration and are either delays or early arrivals. We use distribution rules, a kind of association rules that may have continuous distributions on the consequent. Distribution rules allow the systematic identification of particular conditions, which we call contexts, under which the distribution of trip time deviations differs significantly from the overall deviation distribution. After identifying specific causes of delay the bus company operational managers can make adjustments to the timetables increasing punctuality without disrupting the service. © Springer-Verlag Berlin Heidelberg 2012.

Abstract
Despite their importance, hierarchical clustering has been little explored for semi-supervised algorithms. In this paper, we address the problem of semi-supervised hierarchical clustering by using an active learning solution with cluster-level constraints. This active learning approach is based on a new concept of merge confidence in agglomerative clustering. When there is low confidence in a cluster merge the user is queried and provides a cluster-level constraint. The proposed method is compared with an unsupervised algorithm (average-link) and two state-of-the-art semi-supervised algorithms (pairwise constraints and Constrained Complete-Link). Results show that our algorithm tends to be better than the two semi-supervised algorithms and can achieve a significant improvement when compared to the unsupervised algorithm. Our approach is particularly useful when the number of clusters is high which is the case in many real problems. © 2012 Springer-Verlag Berlin Heidelberg.

Abstract
In the last few years, a huge amount of temporal written information has become widely available on the Internet with the advent of forums, blogs and social networks. This gave rise to a new challenging problem called future retrieval, which consists of extracting future temporal information, that is known in advance, from web sources in order to answer queries that combine text of a future temporal nature. This paper aims to confirm whether web snippets can be used to form an intelligent web that can detect future expected events when their dates are already known. Moreover, the objective is to identify the nature of future texts and understand how these temporal features affect the classification and clustering of the different types of future-related texts: informative texts, scheduled texts and rumor texts. We have conducted a set of comprehensive experiments and the results show that web documents are a valuable source of future data that can be particularly useful in identifying and understanding the future temporal nature of a given implicit temporal query.

Abstract
Collecting and annotating exemplary cases is a costly and critical task that is required in early stages of any classification process. Reducing labeling cost without degrading accuracy calls for a compromise solution which may be achieved with active learning. Common active learning approaches focus on accuracy and assume the availability of a pre-labeled set of exemplary cases covering all classes to learn. This assumption does not necessarily hold. In this paper we study the capabilities of a new active learning approach, d-Confidence, in rapidly covering the case space when compared to the traditional active learning confidence criterion, when the representativeness assumption is not met.. Experimental results also show that; d-Confidence reduces the number of queries required to achieve complete class coverage and tends to improve or maintain classification error.