Robin De Mol

Ordinal assessment of data consistency based on regular expressions

In this paper, a novel assessment method for measurement of consistency of individual, text-valued attributes is proposed. The first novelty of this method is that it allows to express a broad range of well-known consistency measurements in a simple, elegant and standardized way. This property is obtained by relying on the standardized framework of regular expressions to support measurement. The key advantage of using such a highly standardized expression syntax, is that knowledge about consistency becomes portable, exchangeable and easy to access. The second novelty of the method, is that it examines the advantages of using a finite and ordinal scale for expression of measurement. These advantages include a high degree of interpretation and efficient calculations both in terms of time and space complexity.

Wrapper induction by XPath alignment

Dealing with a huge quantity of semi-structured documents and the extraction of information therefrom is an important topic that is getting a lot of attention. Methods that allow to accurately define where the data can be found are then pivotal in constructing a robust solution, allowing for imperfections and structural changes in the source material. In this paper we investigate a wrapper induction method that revolves around aligning XPath elements (steps), allowing a user to generalise upon training examples he gives to the data extraction system. The alignment is based on a modification of the well known Levenshtein edit distance. When the training example XPaths have been aligned with each other they are subsequently merged into the path that generalises, as precise as possible, the examples, so it can be used to accurately fetch the required data from the given source material.

Representing Uncertainty Regarding Satisfaction Degrees Using Possibility Distributions

Evaluating flexible criteria on data leads to degrees of satisfaction. If a datum is uncertain, it can be uncertain to which degree it satisfies the criterion. This uncertainty can be modelled using a possibility distribution over the domain of possible degrees of satisfaction. In this work, we discuss the meaningfulness thereof by looking at the semantics of such a representation of the uncertainty. More specifically, it is shown that defuzzification of such a representation, towards usability in (multi-criteria) decision support systems, corresponds to expressing a clear attitude towards uncertainty (optimistic, pessimistic, cautious, etc.)

Evaluating flexible criteria on uncertain data

Abstract Modern information management systems and databases are rapidly becoming better equipped for handling data imperfections. A common imperfection is uncertainty, indicating that a propertys exact value is not known. Ideally, such systems can be queried uniformly using flexible criteria regardless of whether the underlying data are uncertain or not. The result thereof should always be informative and intuitive, and should reflect to what degree the data satisfy the criteria and to which degree this is uncertain. In this work, we present a novel way to evaluate flexible criteria on uncertain data. The result thereof is a distribution of uncertainty over degrees of satisfaction. These so-called suitability distributions are first constructed for possibilistic data. It is shown that they can be used in all scenarios going from regular, crisp criteria on certain data to flexible criteria on uncertain data, and that they seamlessly generalize other alternatives. Importantly, their interpretation is always the same, so they can be used without needing to have prior knowledge regarding the quality of the data. Afterwards their properties and supported operations are given. Next it is shown that they can also be applied more broadly, for example for probabilistic data. Examples illustrate their rich semantics, ease-of-use and broad applicability.

Handling Uncertainty Degrees in the Evaluation of Relevant Opinions Within a Large Group

Social media makes it possible to involve a large group of people to express their opinions, but not all of these opinions are considered to be relevant from a decision-maker’s point of view. Our approach splits a large group of opinions into clusters—here, a cluster represents a group of similar opinions over a criterion specification. Then, these clusters are categorized as more or less relevant taking into account the decision maker’s point of view over some characteristics—like the level of togetherness (or cohesion) among opinions, and their representativeness. However, these characteristics might include some level of uncertainty. Thus, the aim of this paper is to evaluate relevant opinions taking into account any associated uncertainty. What follows is to produce a meaningful selection of clusters based on their evaluation and their uncertainty degree. Finally, this proposal includes the steps describing the process through an illustrative example.

Applying Suitability Distributions in a Geological Context

Some industrial purposes require specific marine resources. Companies rely on information from resource models to decide where to go and what the cost will be to perform the required extractions. Such models, however, are typical examples of imprecise data sets wherein most data is estimated rather than measured. This is especially true for marine resource models, for which acquiring real data samples is a long and costly endeavor. Consequently, such models are largely computed by interpolating data from a small set of measurements. In this paper, we discuss how we have applied fuzzy set theory on a real data set to deal with these issues. It is further explained how the resulting fuzzy model can be queried so it may be used in a decision making context. To evaluate queries, we use a novel preference modeling and evaluation technique specifically suited for dealing with uncertain data, based on suitability distributions. The technique is illustrated by evaluating an example query and discussing the results.

Data Quality Assessment in Volunteered Geographic Decision Support

Geographic decision support systems aim to integrate and process data originating from different sources and different data providers in order to create suitability models. A suitability model denotes how suitable geographic locations are for a specific purpose on which decision-makers need to make a decision. Particularly in the presence of volunteered information, data quality assessment becomes an important aspect of a decision-making process. Geographic data are commonly prone to incompleteness, imprecision and uncertainty, and this is even more the case with volunteered data. To correctly inform the users, it is essential to communicate not only the suitability degrees highlighted in a suitability model, but also the confidence about these suitability degrees as can be derived from data quality assessment. In this chapter, a novel hierarchical approach for data quality assessment, supporting the computation of associated confidence degrees, is introduced. To illustrate its added value, aspects of the project Transnational and Integrated Long-term marine Exploitation Strategies (TILES) are used. Providing confidence information adds an extra dimension to the decision-making process and leads to more sound decisions.

On the Need for Explicit Confidence Assessments of Flexible Query Answers

Flexible query answering systems aim to exploit data collections in a richer way than traditional systems can do. In approaches where flexible criteria are used to reflect user preferences, expressing query satisfaction becomes a matter of degree. Nowadays, it becomes more and more common that data originating from different sources and different data providers are involved in the processing of a single query. Also, data sets can be very large such that not all data within a database or data store can be trusted to the same extent and consequently the results in a query answer can neither be trusted to the same extent. For this reason, data quality assessment becomes an important aspect of query processing. In this paper we discuss the need for explicit data quality assessments of query results. Indeed, To correctly inform users, it is in our opinion essential to communicate not only the satisfaction degrees in a query answer, but also the confidence about these satisfaction degrees as can be derived from data quality assessment. As illustration, we propose a hierarchical approach for query processing and data quality assessment, supporting the computation of as well a satisfaction degree, as its associated confidence degree for each element of the query result. Providing confidence information adds an extra dimension to query processing and leads to more soundly query answers.

Handling veracity in multi-criteria decision-making : a multi-dimensional approach

Abstract Decision support systems aim to help a decision maker with selecting the option from a set of available options that best meets her or his needs. In multi-criteria based decision support approaches, a suitability degree is computed for each option, reflecting how suitable that option is considering the preferences of the decision maker. Nowadays, it becomes more and more common that data of different quality, originating from different data sets and different data providers have to be integrated and processed in order to compute the suitability degrees. Also, data sets can be very large such that their data become commonly prone to incompleteness, imprecision and uncertainty. Hence, not all data used for decision making can be trusted to the same extent and consequently, neither the results of computations with such data can be trusted to the same extent. For this reason, data quality assessment becomes an important aspect of a decision making process. To correctly inform the users, it is essential to communicate not only the computed suitability degrees of the available options, but also the confidence about these suitability degrees as can be derived from data quality assessment. In this paper a novel multi-dimensional approach for data quality assessment in multi-criteria decision making, supporting the computation of associated confidence degrees, is presented. Providing confidence information adds an extra dimension to the decision making process and leads to more soundly decisions. The added value of the approach is illustrated with aspects of a geographic decision making process.

Indexing Possibilistic Numerical Data: The Interval B + -tree Approach

When record sets become large, indexing becomes a required technique for speeding up querying. This holds for regular databases, but also for ‘fuzzy’ databases. In this paper we propose a novel indexing technique, supporting the querying of imperfect numerical data. A possibility based relational database setting is considered. Our approach is based on a novel adaptation of a B\(^{+}\)-tree, which is currently still one of the most efficient indexing techniques for databases. The leaf nodes of a B\(^{+}\)-tree are enriched with extra data and an extra tree pointer so that interval data can be stored and handled with them, hence the name Interval B\(^{+}\)-tree (IBPT). An IBPT allows to index possibility distributions using a single index structure, offering almost the same benefits as a B\(^{+}\)-tree. We illustrate how an IBPT index can be used to index fuzzy sets and demonstrate its benefits for supporting ‘fuzzy’ querying of ‘fuzzy’ databases. More specifically, we focus on the handling of elementary query criteria that use the so-called compatibility operator IS, which checks whether stored imperfect data are compatible with user preferences (or not).