Zusammenfassung

Weltweit leiden ca. 10% der Bevölkerung an Migräne. Vor allem das Wetter wird oft als Auslöser genannt – wissenschaftlich nachgewiesen ist ein Zusammenhang jedoch noch nicht. Ziel des Projektes ist es daher, Auslöser von Migräneattacken besser zu verstehen. Dazu wurden in vier Jahren ca. 20 000 Attacken mit den zugehörigen Ortsdaten erfasst. Erste Analysen belegen, dass die Häufigkeit von Attacken statistisch signifikant zunimmt, wenn es in den drei Tagen vor einer Attacke zu Änderungen der absoluten Temperatur kam. In der ersten Ausbaustufe des Projektes gab es jedoch noch einige Limitationen. Da die Teilnahme vollständig anonym stattfand, war es nicht möglich, differenzierte Auslösemuster in Subgruppen von Patienten zu erkennen. Dies zu untersuchen, ist ein Ziel des Folgeprojektes. Durch eine anonyme Registrierung der Patienten sollen genauere und individualisierte Analysen möglich sein.

Ein verbessertes Verständnis von Migräne- triggern kann positive Auswirkungen auf die Edukation von Patienten haben, hierzu ist eine individuelle Analyse der Situation des Patienten hilfreich. Aktives Attackenmanagement könnte das bei Migräne-Patienten häufige Meiden von Triggern ablösen und zum erfolgreicheren Umgang mit der Erkrankung führen.

Summary

Approximately 10% of the human population suffers from migraine. In particular, weather is one of the most prominent triggers. However, a correlation between weather and migraine attacks is not proven yet. For this reason, our project aims to get a better understanding of migraine attack triggers. In order to achieve this, we collected about 20 000 attacks with corresponding location information within the last four years. First analyses show a significant increase in migraine frequency in the case of a change in absolute temperature within the three days before an attack. In the first phase of the project were still a few limitations. Since the participation of users was completely anonymous, the detection of differentiated trigger patterns in subgroups of patients was not possible. Therefore, one aim of our follow-up project is the search for such patterns. Allowing patients to register anonymously will lead to more precise and individualized analyses. Furthermore, a better understanding in migraine triggers may have a positive effect on patient education.

For this, an individualized analysis of the patients’ situation will be quite helpful. An active attack management may be able to replace the systematic avoidance of trigger factors, practiced by migraine patients, and thus can lead to a successful handling of the disease.

Mit der zunehmenden Menge textueller Daten im Web 2.0 wächst auch die Notwendigkeit der maschinellen Auswertung dieser Daten, beispielsweise um in Texten geäußerte Meinungen aufzuspüren (Opinion Mining). Im vorliegenden Beitrag wird das Aspect-based Opinion Mining – ein Verfahren mit sehr hohem Detaillierungsgrad – für deutschsprachige Texte anhand eines Projekts für die Versicherungswirtschaft vorgestellt. Es wird gezeigt, dass in Bewertungsplattformen geäußerte Meinungen zu Produkten und Services von Versicherungen mit einer Genauigkeit von etwa 90% und einer Vollständigkeit von ca. 80% für positive und ca. 60% für negative Meinungen erkannt werden können.

In this work, we present a system called PoliTwi, which was designed to detect emerging political topics (Top Topics) in Twitter sooner than other standard information channels. The recognized Top Topics are shared via different channels with the wider public. For the analysis, we have collected about 4,000,000 tweets before and during the parliamentary election 2013 in Germany, from April until September 2013. It is shown, that new topics appearing in Twitter can be detected right after their occurrence. Moreover, we have compared our results to Google Trends. We observed that the topics emerged earlier in Twitter than in Google Trends.

Finally, we show how these topics can be used to extend existing knowledge bases (web ontologies or semantic networks) which are required for concept-level sentiment analysis. For this, we utilized special Twitter hashtags, called sentiment hashtags, used by the German community during the parliamentary election.

In this paper, we discuss the work in progress of our current project focusing on opinion mining in the field of insurance business. The main purpose of this project is to improve Opinion Mining methods for the German language and optimize them with special regard to the insurance business. These improved methods make it possible to extract opinions from user-generated texts (in the insurance domain) in a better quality than today. We fetch the required text data for this study from a huge online community website for customers. There, we find a sufficient number of user reviews about insurance companies, which is necessary for our research. Besides the main purpose of this study, another aim is the development of a prototype. This could then be used to monitor the current "crowd’s opinion" about insurance products and services. For this reason and in order to understand key aspects of the domain, we collaborate with the nobisCum Deutschland GmbH, a German company offering consulting and software development services for the insurance industry. Using one data source and limited evaluation sets, we obtained first results which look promising.

We present a new phrase-based generated list of opinion bearing words and phrases for the German language. The list contains adjectives and nouns as well as adjectiveand noun-based phrases and their opinion values on a continuous range between -1 and +1. For each word or phrase two additional quality measures are given. The list was produced using a large number of product review titles providing a textual assessment and numerical star ratings from Amazon.de. As both, review titles and star ratings, can be regarded as a summary of the writers opinion concerning a product, they are strongly correlated. Thus, the opinion value for a given word or phrase is derived from the mean star rating of review titles which contain the word or phrase. The paper describes the calculation of the opinion values and the corrections which were necessary due to the so-called “J-shaped distribution” of online reviews. The opinion values obtained are amazingly accurate.

Many factors trigger migraine attacks. Weather is often reported to be one of the most common migraine triggers. However, there is little scientific evidence about the underlying mechanisms and causes. In our pilot study, we used smartphone apps and a web form to collect around 4,700 migraine messages in Germany between June 2011 and February 2012. Taking interdiurnal temperature changes as an indicator for changes in the prevailing meteorological conditions, our analyses were focused on the relationship between temperature changes and the frequency of occurrence of migraine attacks. Linear trends were fitted to the total number of migraine messages with respect to temperature changes. Statistical and systematic errors were estimated. Both increases and decreases in temperature lead to a significant increase in the number of migraine messages. A temperature increase (decrease) of 5 °C resulted in an increase of 19 ± 7 % (24 ± 8 %) in the number of migraine messages.

In this paper we present an approach to generate lists of opinion bearing phrases with their opinion values in a continuous range between -- 1 and 1. Opinion phrases that are considered include single adjectives as well as adjective-based phrases with an arbitrary number of words. The opinion values are derived from user review titles and star ratings, as both can be regarded as summaries of the user's opinion about the product under review. Phrases are organized in trees with the opinion bearing adjective as tree root. For trees with missing branches, opinion values then can be calculated using trees with similar branches but different roots. An example list is produced and compared to existing opinion lists.

This paper discusses the work in progress of the ”Migraine Radar” project. The purpose of the project is to validate or disprove the assumed correlation between migraine attacks and weather conditions, especially weather changes. There have been various medical studies on this topic, but the correlation could not be proved with sufficient statistical significance so far. Furthermore, the results of some of the studies are contradictory. For this study, data from the microblogging platform Twitter will be analyzed. Twitter messages (”tweets”) announcing currently or recently happened migraine attacks are retrieved using the Twitter API (Search-API, REST-API - Standard APIs provided by Twitter to retrieve tweet and user data). Weather data from weather information services are linked to the tweets, using the location information from Twitter. For the German language area, the results will be compared with the results obtained from a set of migraine announcements collected with the help of a web form in the same period of time. First statistics indicate that the number of migraine attacks announced in Twitter exceeds the number of cases in former classical studies by far. The project offers a wide range of possibilities to analyze Twitter messages with regard to migraine attacks. Beside the main purpose, it is also possible to analyze the distribution of migraine attacks over the weekdays or over the seasons. Furthermore an investigation of the spatial distribution of migraine attacks is possible. Instead of weather data, other information can be linked to the migraine sample as well. One example could be air pollution data.