It is a complex but very important task to formalize a clinical practice guideline and model it into a representation that can be executed and utilized automatically, since the automatic execution of guidelines at the point-of-care ensures the best available scientic evidence without interrupting clinical work ow. Before a medical text can be translated into such a model, it is necessary to pre-process the text, so that its contents, i. e. the existing medical concepts, can be identied and described unambiguously in order to ensure a correct interpretation.
Semantic Annotation Systems extract medical concepts from the text of guidelines and map them to concepts from medical terminologies, such as the UMLS®Metathesaurus®[36], which contain important additional information. Due to the ambiguity of free text, the correct and automatic identication of medical concepts and the corresponding mapping generated with the help of these systems will probably never be completely correct. Since medical care is an extremely sensitive discipline, the complete reliability of results is crucial for their usability for further processing, which makes it absolutely necessary for experts to controll these results and to modify them, if necessary. This fact led me to develop an editor for the MetaMap Transfer (MMTx) program [2, 3] that enables experts in medical science to solve this task without requiring special skills in information processing. MapFace was designed to realize an easy way to edit the MetaMap results as well as to provide access to all assigned information by a single "mouse-click" in combination with a clearly arranged visualization of the acquired information.information visualization
This master thesis explores various comparison methods for multivariate time series in the application area of stock markets. The data is usually compared by juxtaposition or by superimposition. But even a large enough difference between the price ranges can make a simple comparison of the data difficult.
Stock market data is not limited to stocks. It also includes stock (market) indices such as the Dow Jones and NASDAQ. Stock indices have their own proprietary unit, which is different to the unit of a stock. Time series which have different units are also called heterogeneous data. To compare heterogeneous data by superimposition multiple y-axes are often used. But in most cases the arrangement of the axes makes comparisons between different variables completely arbitrary and comparisons are often misleading. The visualization pioneer Jacques Bertin has studied this problem. One suggestion is to index the values, which transforms all data into values which reflect the percent change compared to an indexing point. The first part of the research is concerned with the identification of relevant comparison methods for multivariate time series. The major part of the research deals with the evaluation of an advanced comparison method based on indexing. Another important part is the investigation of the importance of the used axis scale. Differences between linear and logarithmic scale are analyzed for effects on user performance. The major part of this work is a comparative study about three visual comparison methods (visualization types) for multivariate time series. The three tested comparison methods are juxtaposition, superimposition and indexing. 24 test persons participated in the study. Each participant had to complete 14 tasks for each one of the three visualization types. The task completion time and the task correctness for every task were measured and later used for statistical analyses. This work further presents state of the art research about other visualizations suited for visual comparison tasks. The prototype application incorporates several common stock market visualizations such as line charts, OHLC charts and candlestick charts. Basic comparison methods like juxtaposition and superimposition are available. A more advanced comparison method based on indexing was implemented. The usability test results support the assumption that the indexing method enables the user to perform comparison tasks with much less estimation errors. The task completion time is not significantly different. The free selection of the indexing point makes comparisons for a certain time period more effective and delivers more precise results. A post-test survey showed that the majority of the participants favor the indexing method over the two other visualization types. The test results for the usage of different scales indicate that tasks were faster completed when using logarithmic scales. The task correctness rate was not significantly different between linear and logarithmic scales.This master thesis explores key requirements for the design of interactive visualization methods for comparing multivariate stock market data.
Superimposed line charts are commonly used for comparison of stocks in modern stock market visualizations. A problem is appearing, when the difference(s) between the price ranges is (are) large. These differences result in distortion and a display of flattened curves, which make comparisons very difficult. Another problem is that stocks and stock indices do not have the same unit. In this case juxtaposed line charts are often used to compare the data. However the data does not share the same y-axis. Other than that, superimposed charts with multiple y-axes might be used. Relationships are then depending on the ranges of the involved y-axes, which makes the visual appearance often completely arbitrary. The third problem is that stock market visualizations often use linear scales. Unfortunately this scale can lead to false conclusion of comparisons with percent estimations. A user & task analysis with six domain experts was conducted to find common user tasks and requirements for the design of a stock market visualization prototype. The main subjects of the analysis are user tasks, visualizations & interactions and related economic data. After the development of the prototype, a user study with five domain experts was conducted. The purpose of this study is to gain feedback about the usability and the applicability of the prototype from expert users. All participants of the user study have agreed that a comparison of multiple stocks by superimposition is more effective than by juxtaposition. The domain experts also think that indexed line charts are an excellent method to compare percent changes. The indexing method can display various stock market data with different units using only one chart. This makes relative comparison tasks of stock market data very effective. The ability to select the indexing point freely is considered by all participants as an important and useful function. According to the user study, the logarithmic scale is more useful than the linear scale for comparisons of stock market data. Also it is best to compare a group of stocks or a stock market index with economic data like GDP or interest rates. This removes any individual effects of a particular stock.Visualization techniques for time-oriented data can be divided into two main categories: dynamic and static visualizations. Dynamic visualizations map the time directly to the time and present the data of different time-points or -intervals successively.
This means the visualization changes its appearance over time. Compared to that, in static visualizations the factor time is mapped to space (e.g. a line chart where the time is projected to the x-axis). Static visualizations can be further differentiated in diagrams which present only states or diagrams also visualizing the development by e.g. some kind of arrows or lines. The aim of this thesis was an empirical comparison between a dynamic and two static visualizations which are suitable for analyzing relationships of two variables and their development over time. The chosen dynamic visualization presents the development of the data with an animated scatter plot. Within the static visualizations ("Small Multiples" and "Trace view") the time intervals and the data recorded in those time intervals are displayed in small pictures arranged side by side. Additionally data points belonging to an object are connected by lines in the "Trace view". Those connections between the data points are continued over all small pictures and therefore visualize the whole development of the parameters explicitly. To make the comparison possible it was necessary to extend the prototype TimeRider by the "Small Multiple" visualization with the possibility to activate the traces resulting in the previously described "Trace view". The quantitative study was carried out over the internet, where each of the 29 participants had to solve tasks with all visualizations. The tasks covered three topics: the examination of trends, the identification and analysis of outliers as well as the identification of clusters and their development over time. The analysis of the results showed that all visualizations caused similar success rates and completion times for tasks concerning outliers and clusters. Regarding data movements within clusters and small changes in the data the animation outperformed the other visualizations significantly in correctness. The results also showed that participants solved tasks concerning the analysis of specific data points and the comparison between them significantly faster and with higher success rates with the dynamic visualization. However for nearly every hypothesis the results were dataset-dependent and had to be analyzed split by dataset which reduces the significance of the statistical results. But allover the results indicate that animation is better or at least equal to the two static visualizations for the analysis of time-oriented data.Abstract Information visualization is of great importance in order to allow users to display complex and large amounts of data. It has a long history on the desktop computer, and therefore there are already numerous tools and frameworks that satisfy the needs of the majority of users. Because of the proliferation of mobile devices in recent years and because of the improvement of said mobile devices with regard to performance and screen quality, the demand for information visualization on mobile devices has increased. Since this is a new area, there are relatively few tools and frameworks for this purpose. The topic of this paper is introduced by displaying the already existing tools and frameworks in Android. Then their functionality and options are briefly explained. Furthermore, the structure of prefuse, one of the most successful desktop frameworks concerning information visualization, is analyzed. In the course of this analysis, a review of the actual source code is added to the information that has been obtained from the State of the Art on prefuse. Particular attention is paid to used design patterns. Finally, the conducted porting and its final result, AndroidPrefuse, is described. In general, the architecture of AndroidPrefuse after porting has remained the same as prefuse. To demonstrate the successful porting, a simple scatterplot in AndroidPrefuse has been implemented or rather ported. In addition, the following questions regarding the porting are answered: Which options for the porting were available? Why were the AWT and Swing libraries adopted for AndroidPrefuse? What challenges have been encountered during porting, and how were they accomplished? Which prefuse parts were completely adopted, and which were only partially or not adopted at all?
efficiently store network structured data. But while plenty of research has gone into the development of this technology, querying for a subset of the data is lacking user friendly interfaces. Of the two main query methods, graph pattern matching and graph traversal, the first has received more attention and more methods providing visual support in querying are available. The latter, graph traversal, while being very powerful, has seen little advances in visual querying. This thesis aims at providing a novel user interface for graph traversal query formulation - a visual query language for graph databases - entitled Pygmalion Query. In the thesis, first a literature review is undertaken to discover any previous approaches taken at such a visual query language. With it, gaps in the currently available research are identified. Following the literature review, needs and requirements are identified from different sources, such as queries posted online and documentation for graph traversal query languages. After selection of minimum required features, the design for Pygmalion Query is created. A web-based implementation, built on available frameworks, is implemented. Following the creation of Pygmalion Query, a twofold evaluation is conducted. An expert review serves as the initial confirmation of the approach taken. Using feedback coming from the experts, an updated implementation is created. A small comparative user study is carried to test for usability. The results of the expert review and user study indicate a positive usability effect of Pygmalion Query in the formulation of graph traversal queries over the currently available solutions. The participants of the study, in greatest part novice users, are more likely to complete the tasks posed to them with the visual query language.
In information visualization set-typed data refers to datasets that represent element-set memberships, such as which countries (sets) produce a certain product (element), or which genres (sets) a movie (element) belongs to. Set-typed data appears in various forms and can serve as a data models in various data analysis scenarios. One of the main challenges in the context of set-typed data visualization is scalability. Traditionally, Euler and Venn diagrams count as two of the most popular set visualizations that depict the concepts from set theory. However, despite the widespread usage of these diagrams across several scientific fields they lack the ability of visualizing more than three sets without becoming too complex. This limits their applicability to data analysis scenarios that involve hundreds of sets such as the world countries. Besides Euler and Venn diagrams a variety of visualization techniques for set-typed data has been developed over the past. Typically, existing techniques scale well with either an increasing number of elements or an increasing number of sets. The goal of this thesis is to develop a set visualization technique which offers high scalability in both the number of sets and elements. The proposed technique, called Scets, employs different aggregations of set-type data, and uses a matrix layout to visualize the aggregated information. Furthermore, it allows users to explore the aggregated information interactively. The implemented prototype uses modern web technologies which make the visualization both able to handle a large amount of data using server-side backend, and accessible to a wide range of users using web-based frontend. Two different use cases demonstrate how the proposed visualization technique helps to investigate real-world data and enables users in an intuitive way to reveal several patterns which could not be easily detected by other visualization techniques.
Information present in resumes has temporal as well as spatial dimensions, such as a person¿s work experience or education. This information by itself is comprehensible - it is very clear when, where, and for how long a person has worked or studied. It becomes increasingly difficult to maintain a clear overview when comparing multiple resumes, specifically their chronological information. Navigating back and forth between resumes, switching between views and attempting to compare every single event is highly inefficient. The most common and intuitive way to compare events from multiple resumes, would be to view them in a side-by-side fashion. Many similarities can be quickly recognized by doing this, but in cases where more than two resumes need to be compared simultaneously, the overview becomes cluttered, the amount of information increases substantially and comparing events becomes a difficult task. We propose the design and implementation of a web application - CV3 - that is capable of comparing multiple chronologies and visualizing the output in a clear manner, whilst maintaining a clean overview. Our approach supports users in filtering timeline events across all resumes, recognizing and extracting relevant information from resumes and visualizing their similarities efficiently in a single overview.
Information retrieved from the real world often contains some kind of inherent uncertainty. In recent years there has been an effort to incorporate this aspect of data into visualizations. This is also true for the representation of temporal data. How this can be done in an intuitive way is still one of many open questions regarding temporal uncertainty visualization. This thesis presents two subsequent user studies, which aim at providing insights into this matter. In the first study, called the Drawing Study, 32 participants are asked to draw their own visualization designs, based on predefined scenarios and tasks. These drawings are analyzed through an open coding approach. The analysis yields a list of hypotheses regarding the intuitiveness of temporal uncertainty visualization. From this list a selection of hypotheses leads to more concrete research questions, which form the basis for the second study, called the User Survey. In this online survey 60 participants compare and rate different visualization approaches in several scenarios. This rating of intuitiveness yields valuable insights for future visualization design. The results indicate that icon representations are not considered intuitive, even though they might seem to be at first glance. It can be argued that icons just need to be designed specifically for certain tasks and scenarios to be perceived intuitive. Furthermore, the results suggest that most people prefer to have uncertainty presented to them, even if it is not relevant for the task at hand. This finding could have important implications for the design of future visualizations.
As companies gather more and more data, we need to find a way to allow interested decision makers to access this data in an efficient way. In the context of sports practice, users could benefit from suggestions about new sports they could try out and the company could increase its sales. This work aims to support the analysts, simultaneously domain experts and IT laymen, in their data exploration and suggestion retrieval tasks through a user friendly interface, abstracting away the complexity of formulating expressive queries into the visual domain. We present a characterization and task analysis for this domain, and a prototype that meets the requirements emerging from them, based on an interdisciplinary literature research. The resulting prototype combines a visual query language with a collaborative filtering approach to render suggestions for new activities, and show multiple types of relationships in a visually compelling way. It has been implemented as a web application that handles the transformation of user input from a graphical pattern into a database query language and the results of this query into an easy to digest information representation. We conclude with an expert interview to validate the design for analysis and exploration.