Matthias Keller

CSS Code Quality: A Metric for Abstractness; Or Why Humans Beat Machines in CSS Coding

Authoring CSS is a complex, time consuming task requiring not only skilled human graphic designers but also skilled human coders. Practice shows that today human authored code is still superior to machine generated CSS, but the code characteristics which make the difference have not been researched or even quantified yet. In this paper we introduce the abstractness factor, a quality metric which reveals the advantages of human authored code and can serve as an optimization criterion and benchmark for automated CSS coding. We argue that a high abstractness factor represents a high maintainability and reusability of the presentation document as well as the content document. By an evaluation of 100,000 HTML pages randomly gathered from the Web we show that today’s typical style sheet document has a significantly higher abstractness factor compared to code fully machine generated by state-of-the-art applications.

Cascading style sheets: a novel approach towards productive styling with today's standards

In this paper we present an approach of generating Cascading Style Sheet documents automatically if the desired effect on the content elements is specified. While a Web user agent resolves the CSS rules and computes their effect, our approach handles the way back. We argue, that this can remarkably improve CSS productivity, since the process of CSS authoring always involves this direction implicitly. Our approach claims a new and innovative way to reuse chunks of markup together with its presentation. It furthermore bears potential for the optimization and reorganization of CSS documents. We describe criteria for CSS code quality we oriented on, including a quantitative indicator for the abstractness of a CSS presentation specification. An evaluation and recomputation of the CSS for 25.000 HTML documents shows that concerning these criteria the automatically generated code comes close to manually authored code.

Fuzzy decision tree to classify complex fractionated atrial electrograms

Abstract Catheter ablation has emerged as an effective treatment strategy for atrial fibrillation (AF) in recent years. During AF, complex fractionated atrial electrograms (CFAE) can be recorded and are known to be a potential target for ablation. Automatic algorithms have been developed to simplify CFAE detection, but they are often based on a single descriptor or a set of descriptors in combination with sharp decision classifiers. However, these methods do not reflect the progressive transition between CFAE classes. The aim of this study was to develop an automatic classification algorithm, which combines the information of a complete set of descriptors and allows for progressive and transparent decisions. We designed a method to automatically analyze CFAE based on a set of descriptors representing various aspects, such as shape, amplitude and temporal characteristics. A fuzzy decision tree (FDT) was trained and evaluated on 429 predefined electrograms. CFAE were classified into four subgroups with a correct rate of 81±3%. Electrograms with continuous activity were detected with a correct rate of 100%. In addition, a percentage of certainty is given for each electrogram to enable a comprehensive and transparent decision. The proposed FDT is able to classify CFAE with respect to their progressive transition and may allow objective and reproducible CFAE interpretation for clinical use.

Characterization of Radiofrequency Ablation Lesion Development Based on Simulated and Measured Intracardiac Electrograms

Radiofrequency ablation (RFA) therapy is the gold standard in interventional treatment of many cardiac arrhythmias. A major obstacle is nontransmural lesions, leading to recurrence of arrhythmias. Recent clinical studies have suggested intracardiac electrogram (EGM) criteria as a promising marker to evaluate lesion development. Seeking for a deeper understanding of underlying mechanisms, we established a simulation approach for acute RFA lesions. Ablation lesions were modeled by a passive necrotic core surrounded by a borderzone with properties of heated myocardium. Herein, conduction velocity and electrophysiological properties were altered. We simulated EGMs during RFA to study the relation between lesion formation and EGM changes using the bidomain model. Simulations were performed on a three-dimensional setup including a geometrically detailed representation of the catheter with highly conductive electrodes. For validation, EGMs recorded during RFA procedures in five patients were analyzed and compared to simulation results. Clinical data showed major changes in the distal unipolar EGM. During RFA, the negative peak amplitude decreased up to 104% and maximum negative deflection was up to 88% smaller at the end of the ablation sequence. These changes mainly occurred in the first 10 s after ablation onset. Simulated unipolar EGMs reproduced the clinical changes, reaching up to 83% negative peak amplitude reduction and 80% decrease in maximum negative deflection for transmural lesions. In future studies, the established model may enable the development of further EGM criteria for transmural lesions even for complex geometries in order to support clinical therapy.

Beyond the Web Graph: Mining the Information Architecture of the WWW with Navigation Structure Graphs

Large Web sites contain a plethora of different menus and navigation aids, which implement systems of content organization as hierarchies, linear structures or matrices. Humans are able to decode the fine-grained content organization because they are aware of the different access methods provided by navigation systems and understand the higher-level information architecture. In contrast, current methods of link analysis cannot extract such a detailed model of the information architecture and are not able to recognize site boundaries and content hierarchies the way humans do. In this paper present a new approach of mining navigation systems that increases the precision of Web structure mining. Instead of analyzing the complete Web graph spanned by pages and hyperlinks, sub graphs called Navigation Structure Graphs (NSGs) are analyzed. A NSG represents the hyperlinks belonging to a certain navigation system. We demonstrate the capabilities of NSGs for analyzing the organization of Web sites and present our research on mining NSGs.

Characterization of Complex Fractionated Atrial Electrograms by Sample Entropy: An International Multi-Center Study

Atrial fibrillation (AF) is the most commonly clinically-encountered arrhythmia. Catheter ablation of AF is mainly based on trigger elimination and modification of the AF substrate. Substrate mapping ablation of complex fractionated atrial electrograms (CFAEs) has emerged to be a promising technique. To improve substrate mapping based on CFAE analysis, automatic detection algorithms need to be developed in order to simplify and accelerate the ablation procedures. According to the latest studies, the level of fractionation has been shown to be promisingly well estimated from CFAE measured during radio frequency (RF) ablation of AF. The nature of CFAE is generally nonlinear and nonstationary, so the use of complexity measures is considered to be the appropriate technique for the analysis of AF records. This work proposes the use of sample entropy (SampEn), not only as a way to discern between non-fractionated and fractionated atrial electrograms (A-EGM), Entropy 2015, 17 7494 but also as a tool for characterizing the degree of A-EGM regularity, which is linked to changes in the AF substrate and to heart tissue damage. The use of SampEn combined with a blind parameter estimation optimization process enables the classification between CFAE and non-CFAE with statistical significance (p < 0:001), 0.89 area under the ROC, 86% specificity and 77% sensitivity over a mixed database of A-EGM combined from two independent CFAE signal databases, recorded during RF ablation of AF in two EU countries (542 signals in total). On the basis of the results obtained in this study, it can be suggested that the use of SampEn is suitable for real-time support during navigation of RF ablation of AF, as only 1.5 seconds of signal segments need to be analyzed.

GRABEX: A Graph-Based Method for Web Site Block Classification and Its Application on Mining Breadcrumb Trails

In order to interact with a Web site, humans must be able to distinguish and understand the purposes of different page blocks, e.g. header, navigation bar or content area. In case of navigational blocks, the block type determines the functionality of the hyperlinks it contains. For example, the hyperlinks in the main menu block represent the main topics of a site while the hyperlinks in a breadcrumb trail show the location in the content hierarchy. Hence, mining navigational blocks of specific types can provide valuable input for applications in the fields of crawling, ranking or presenting search results. However, analyzing visual features in order to identify specific navigational blocks as humans do is a difficult, resource-consuming task and a general solution does not exist yet. In this paper, we propose a novel approach to the problem and present the Graph-based block extraction method (GRABEX) that can be adapted to classify different types of navigational blocks. The fundamental concept is that a separate graph-based link-analysis is conducted for groups of blocks. Each block group consists of blocks from different pages that have similar CSS class attributes. This allows discovering navigational blocks of specific types, e.g. breadcrumb trails, without analyzing any presentational features. We apply our method to mine breadcrumb trails and are the first to describe an applicable solution to this problem. In an extensive evaluation including 700 different sites, the GRABEX-method performed with perfect precision and high recall.

Search result presentation: supporting post-search navigation by integration of taxonomy data

As a result of additional semantic annotations and novel mining methods, Web site taxonomies are more and more available to machines, including search engines. Recent research shows that after a search result is clicked, users often continue navigating on the destination site because in many cases a single document cannot satisfy the information need. The role Web site taxonomies play in this post-search navigation phase has not yet been researched. In this paper we analyze in an empirical study of three highly-frequented Web sites how Web site taxonomies influence the next browsing steps of users arriving from a search engine. The study reveals that users not randomly explore the destination site, but proceed to the direct child nodes of the landing page with significantly higher frequency compared to the other linked pages. We conclude that the common post-search navigation strategy in taxonomies is to descend towards more specific results. The study has interesting implications for the presentation of search results. Current search engines focus on summarizing the linked document only. In doing so, search engines ignore the fact the linked documents are in many cases just the starting point for further navigation. Based on the observed post-search navigation strategy, we propose to include information about child nodes of linked documents in the presentation of search results. Users would benefit by saving clicks, because they could not only estimate whether the linked document provides useful information, but also whether post-search navigation is promising.

Comparison of simulated and clinical intracardiac electrograms

Intracardiac electrograms are the key in understanding, interpretation and treatment of cardiac arrhythmias. However, electrogram morphologies are strongly variable due to catheter position, orientation and contact. Simulations of intracardiac electrograms can improve comprehension and quantification of influencing parameters and therefore reduce misinterpretations. In this study simulated intracardiac electrograms are analyzed regarding tilt angles of the catheter relative to the propagation direction, electrode tissue distances as well as clinical filter settings. Catheter signals are computed on a realistic 3D catheter geometry using bidomain simulations of cardiac electrophysiology. Thereby high conductivities of the catheter electrodes are taken into account. For validation, simulated electrograms are compared with in vivo electrograms recorded during an EP-study with direct annotation of catheter orientation and tissue contact. Good agreement was reached regarding timing and signal width of simulated and measured electrograms. Correlation was 0.92±0.07 for bipolar, 0.92±0.05 for unipolar distal and 0.80 ± 0.12 for unipolar proximal electrograms for different catheter orientations and locations.

Mining taxonomies from web menus: rule-based concepts and algorithms

The logical hierarchies of Web sites (i.e. Web site taxonomies) are obvious to humans, because humans can distinguish different menu levels and their relationships. But such accurate information about the logical structure is not yet available to machines. Many applications would benefit if Web site taxonomies could be mined from menus, but it was an almost unsolvable problem in the past. While a tag newly introduced in HTML5 and novel mining methods allow to distinguish menus from other contents today, it has not yet been researched, how the underlying taxonomies can be extracted, given the menus. In this paper we present the first detailed analysis of the problem and introduce rule-based concepts for addressing each identified sub problem. We report on a large-scale study on mining hierarchical menus of 350 randomly selected domains. Our methods allow extracting Web site taxonomy information that was not available before with high precision and high recall.