Panagiotis Stamatopoulos

A Memory-Based Approach to Anti-Spam Filtering for Mailing Lists

This paper presents an extensive empirical evaluation of memory-based learning in the context of anti-spam filtering, a novel cost-sensitive application of text categorization that attempts to identify automatically unsolicited commercial messages that flood mailboxes. Focusing on anti-spam filtering for mailing lists, a thorough investigation of the effectiveness of a memory-based anti-spam filter is performed using a publicly available corpus. The investigation includes different attribute and distance-weighting schemes, and studies on the effect of the neighborhood size, the size of the attribute set, and the size of the training corpus. Three different cost scenarios are identified, and suitable cost-sensitive evaluation functions are employed. We conclude that memory-based anti-spam filtering for mailing lists is practically feasible, especially when combined with additional safety nets. Compared to a previously tested Naive Bayes filter, the memory-based filter performs on average better, particularly when the misclassification cost for non-spam messages is high.

Summarization from medical documents: a survey

OBJECTIVE The aim of this paper is to survey the recent work in medical documents summarization. BACKGROUND During the last decade, documents summarization got increasing attention by the AI research community. More recently it also attracted the interest of the medical research community as well, due to the enormous growth of information that is available to the physicians and researchers in medicine, through the large and growing number of published journals, conference proceedings, medical sites and portals on the World Wide Web, electronic medical records, etc. METHODOLOGY This survey gives first a general background on documents summarization, presenting the factors that summarization depends upon, discussing evaluation issues and describing briefly the various types of summarization techniques. It then examines the characteristics of the medical domain through the different types of medical documents. Finally, it presents and discusses the summarization techniques used so far in the medical domain, referring to the corresponding systems and their characteristics. DISCUSSION AND CONCLUSIONS The paper discusses thoroughly the promising paths for future research in medical documents summarization. It mainly focuses on the issue of scaling to large collections of documents in various languages and from different media, on personalization issues, on portability to new sub-domains, and on the integration of summarization technology in practical applications.

Crew Assignment via Constraint Programming: Integrating Column Generation and Heuristic Tree Search

The Airline Crew Assignment Problem (ACA) consists of assigning lines of work to a set of crew members such that a set of activities is partitioned and the costs for that assignment are minimized. Especially for European airline companies, complex constraints defining the feasibility of a line of work have to be respected. We developed two different algorithms to tackle the large scale optimization problem of Airline Crew Assignment. The first is an application of the Constraint Programming (CP) based Column Generation Framework. The second approach performs a CP based heuristic tree search. We present how both algorithms can be coupled to overcome their inherent weaknesses by integrating methods from Constraint Programming and Operations Research. Numerical results show the superiority of the hybrid algorithm in comparison to CP based tree search and column generation alone.

A Generic Object-Oriented Constraint-Based Model for University Course Timetabling

The construction of course timetables for academic institutions is a very difficult problem with a lot of constraints that have to be respected and a huge search space to be explored, even if the size of the problem input is not significantly large, due to the exponential number of the possible feasible timetables. On the other hand, the problem itself does not have a widely approved definition, since different variations of it are faced by different departments. However, there exists a set of entities and constraints among them which are common to every possible instantiation of the timetabling problem. In this paper, we present a model of this common core in terms of ILOG SOLVER, a constraint programming object-oriented C++ library, and we show the way this model may be extended to cover the needs of a specific academic unit.

Crew Pairing Optimization with Genetic Algorithms

We present an algorithm for the crew pairing problem, an optimization problem that is part of the airline crew scheduling procedure. A pairing is a round trip starting and ending at the home base, which is susceptible to constraints that arise due to laws and regulations. The purpose of the crew pairing problem is to generate a set of pairings with minimal cost, covering all flight legs that the company has to carry out during a predefined time period. The proposed solution is a two-phase procedure. For the first phase, the pairing generation, a depth first search approach is employed. The second phase deals with the selection of a subset of the generated pairings with near optimal cost. This problem, which is modelled by a set covering formulation, is solved with a genetic algorithm. The presented method was tested on actual flight data of Olympic Airways.

Combinatorial optimization through statistical instance-based learning

Different successful heuristic approaches have been proposed for solving combinatorial optimization problems. Commonly, each of them is specialized to serve a different purpose or address specific difficulties. However, most combinatorial problems that model real world applications have a priori well known measurable properties. Embedded machine learning methods may aid towards the recognition and utilization of these properties for the achievement of satisfactory solutions. In this paper, we present a heuristic methodology which employs the instance-based machine learning paradigm. This methodology can be adequately configured for several types of optimization problems which are known to have certain properties. Experimental results are discussed concerning two well known problems, namely the knapsack problem and the set partitioning problem. These results show that the proposed approach is able to find significantly better solutions compared to intuitive search methods based on heuristics which are usually applied to the specific problems.

NEARLY OPTIMUM TIMETABLE CONSTRUCTION THROUGH CLP AND INTELLIGENT SEARCH

The course timetable construction is a procedure that every academic department has to carry out at least twice annually, more times if some of the requirements change. These requirements indicate that a collection of elements must be taken in mind in order for an acceptable solution to be found. They come either from the inherent constraints of the problem or from the involved parties, namely teachers and students. A requirement that is very difficult to statisfy is the one of optimality, which means that the constructed timetable should be the best among the legal ones, according to some quantified quality criteria. In this paper, a method for tackling the course timetable construction problem for academic departments is presented, which is based on Constraint Logic Programming (CLP) for the early pruning of the search space and on the usage of intelligent heuristics in order to guide the search to the generation of nearly optimum solutions. A specific system is presented, named ACTS (Automated Course Timetabling System), which has been implemented in the ECLiPSe language. This system is currently in use by the Department of Informatics of the University of Athens for the purpose of aiding the semester course timetable construction.

Using synchronic and diachronic relations for summarizing multiple documents describing evolving events

In this paper we present a fresh look at the problem of summarizing evolving events from multiple sources. After a discussion concerning the nature of evolving events we introduce a distinction between linearly and non-linearly evolving events. We present then a general methodology for the automatic creation of summaries from evolving events. At its heart lie the notions of Synchronic and Diachronic cross-document Relations (SDRs), whose aim is the identification of similarities and differences between sources, from a synchronical and diachronical perspective. SDRs do not connect documents or textual elements found therein, but structures one might call messages. Applying this methodology will yield a set of messages and relations, SDRs, connecting them, that is a graph which we call grid. We will show how such a grid can be considered as the starting point of a Natural Language Generation System. The methodology is evaluated in two case-studies, one for linearly evolving events (descriptions of football matches) and another one for non-linearly evolving events (terrorist incidents involving hostages). In both cases we evaluate the results produced by our computational systems.

Navigation Pattern Discovery Using Grammatical Inference

We present a method for modeling user navigation on a web site using grammatical inference of stochastic regular grammars. With this method we achieve better models than the previously used first order Markov chains, in terms of predictive accuracy and utility of recommendations. In order to obtain comparable results, we apply the same grammatical inference algorithms on Markov chains, modeled as probabilistic automata. The automata induced in this way perform better than the original Markov chains, as models for user navigation, but they are considerably inferior to the automata induced by the traditional grammatical inference methods. The evaluation of our method was based on two web usage data sets from two very dissimilar web sites. It consisted in producing, for each user, a personalized list of recommendations and then measuring its recall and expected utility.

Construction and Repair: A Hybrid Approach to Search in CSPs

In order to obtain a solution to a constraint satisfaction problem, constructive methods iteratively extend a consistent partial assignment until all problem variables are instantiated. If the current partial assignment is proved to be inconsistent, it is then necessary to backtrack and perform alternative instantiations. On the other hand, reparative methods iteratively repair an inconsistent complete assignment until it becomes consistent. In this research, we investigate an approach which allows for the combination of constructive and reparative methods, in the hope of exploiting their intrinsic advantages and circumventing their shortcomings. Initially, we discuss a general hybrid method called cr and then proceed to specify its parameters in order to provide a fully operational search method called cnr. The reparative stage therein is of particular interest: we employ techniques borrowed from local search and propose a general cost function for evaluating partial assignments. In addition, we present experimental results on the open-shop scheduling problem. The new method is compared against specialized algorithms and exhibits outstanding performance, yielding solutions of high quality and even improving the best known solution to a number of instances.