Explicit semantic enrichment make digital scholarly publications potentially easy to find, to navigate, to organize and to understand. But whereas the generation of explicit semantic information is common in fields like biomedical research, comparable approaches are rare for domains in the humanities. Apart from a lack of authoritative structured knowledge bases formalizing the respective conceptualizations and terminologies, many experts from specialized fields of research seem reluctant to employ the technologies and methods that are currently available for the generation of structured knowledge representations. However, human involvement is indispensable in the organization and application of the domain-specific knowledge representations necessary for the contextualization of structured semantic data extracted from textual and scholarly resources. Over the past decade, various efforts have been made towards openly accessible online knowledge graphs containing collaboratively edited, structured and cross-linked data. Such public knowledge bases might be suitable as a starting point for defining formalized domain knowledge representations, with which the subjects and findings of a research domain can be described. Extensive re-use of the widely adopted shared conceptualizations from a large collaborative knowledge base could be in more than one way beneficial to processes of semantic enrichment, especially those involving domain experts with less-technical backgrounds.Weniger anzeigen
Crowd ideation platforms provide a promising approach for supporting the idea generation process. Research has shown that presenting a set of similar or diverse ideas to users during ideation leads them to come-up with more creative ideas. In this paper, we describe Innovonto, a crowd ideation platform that leverages semantic web technologies and human collaboration to identify similar and diverse ideas in order to enhance the creativity of generated ideas. Therefore, the approach implemented captures first the conceptualization of users' ideas. Then, a matching system is employed to compute similarities between all ideas in near real time. Furthermore, this technical report outlines the results obtained from the evaluation of Innovonto platform. The hallway study, conducted at our research group, allowed us to test each step of Innovonto platform as well as the proposed approach in assessing similarities between ideas. As results, we received 20 ideas and 23 feedbacks from 9 users. The analysis of the results shows good performance of Innovonto steps and confirms the findings of existing research.Weniger anzeigen
Genius Media Group Incorporated is a collaborative annotation platform and was founded by Mahbod Moghadam, Tom Lehman and Ilan Zechory with the aim to annotate lyrics, which have no license and can be interpreted in form of in- line annotation by members. The first version was launched October 2009 as Rap Exegesis, then it changed to Rap Genius in December 2009 and finally in July 2014 the title Genius was given. Genius members have six different roles that are closely tied to authorizations sequentially: Whitehat, Editor, Moderator, Verified Artist, Mediator and Staff. We monitored Genius activities on firehose for five weeks, collected 1.3 million activities, 762 thousand of them are annotation activities1. We registered 57 thousand unique users and found, that users generate on average 13.33 annotation activities in this period of analysis, which is 0.36 annotation activities per day. The distribution over user groups displays the roles Moderator, Staff, Artist, Mediator and Whitehat. Whitehat embodies the most registered user, but when it comes to drive Genius ahead, then those roles are presented in the following sequence: Artist, Staff, Mediator, Moderator, Editor and at the end is the role Whitehat. Intelligence Quotient (IQ) can be earned by the most of activities and indicate experience of a member. Although a count of IQs is required to do certain activities, for instance to post into forums, but it does not promote automatically a member’s role to become a higher member level. High-quality annotations and decision maker such as an Editor establish nomination criteria. Earning IQs implies to edit pages; a page is edited on average 295 times, which varies greatly from the me- dian (195) times. This indicates that some pages attract users more than others. For developers Genius provides API, documentation and support forum as well as there are sub- domains in different countries and languages. We attempt to discover members' collaboration by editing Genius pages and for this purpose we clarify the social, technical and participation architecture of Genius, such as member’s permissions as well as options, activity types and distribution of page edits. The following technical report is structured as follows: Section 2 introduces the social structure of Genius, how to interact with the user interface, being a member and the relation between member roles, annotation and earned IQs. Section 3 continues with the technical structure, in which Genius subdomains are presented, what technical options are there for developers to bind Genius services in applications, firehose as notifications process as well as demographic trends of users at Genius. Section 4 describes our member activities study on Genius and which new findings we determined. Finally, section 5 presents our conclusions.Weniger anzeigen
A primary goal of a research infrastructure for data management should be to enable efficient data discovery and integration of heterogeneous data. The German Federation for Biological Data (GFBio) was envisioned by this goal. The basic component, that enables such interoperability and serves as a backbone for such a platform, is the GFBio Terminology Service (GFBio TS). It acts as a semantic platform for accessing, developing and reasoning over terminological resources within the biological and environmental domain. A RESTful API gives access to these terminological resources in a uniform way regardless of their degree of complexity and whether they are internally stored or externally accessed through their web services. Additionally, a set of widgets with an intrinsic API connection are made available for an easy integration in applications and web interfaces. Based on the requirements of the GFBio partners, we describe the added value that is provided by the GFBio Terminology Service with practical scenarios but also, what challenges we still face. We conclude by describing our current activities and future developments.Weniger anzeigen
Higher Order Demand Propagation as proposed in [Pa98] provides a non-standard denotational semantics for a realistic functional language. This semantics can be used to deduce generalised strictness information for higher order polymorphic functions. This report provides the formal proof for the correctness of this strictness information with respect to the non-strict standard semantics.
Vor sechzig Jahren wurde der mechanische Speicher der Rechenmaschine Z1 fertiggestellt. Konrad Zuses Z1 und Z3, zwischen 1936 und 1941 gebaut, bestechen noch heute durch ihre Eleganz. Beide Maschinen hatten denselben logischen Aufbau und waren die ersten vollautomatischen, programmgesteuerten Rechenmaschinen der Welt. Das Papier beschreibt die Architektur von beiden Maschinen.
We address the problem of how to cover a set of "required points" by a small number of "axis-parallel ellipses" that avoid a second set of "forbidden points". We study geometric properties of such covers and present an efficient randomized approximation algorithm for the cover construction. This question is motivated by a special pattern recognition task where one has to identify ellipse-shaped protein spots in two-dimensional electrophoresis images.Weniger anzeigen
Color-segmentation is very sensitive to changes in the intensity of light. Many algorithms do not tolerate variations in color hue which correspond, in fact, to the same object. Learning Vector Quantization (LVQ) networks learn to recognize groups of similar input vectors in such a way that neurons physically near to each other in the neuron layer respond to similar input vectors. Learning is supervised, the inputs vectors into target classes are chosen by the user. In this work a new algorithm based on LVQ is presented. It involves neural networks that operate directly on the image pixels with a decision function. This algorithm has been applied to spotting and tracking human faces, and shows more robustness than other algorithms for the same task.Weniger anzeigen
The report deals with the largely unexplored field of ontology-driven, knowledge based trend detection by means of text mining, focusing on the development of trend ontologies. The difficulties of trend detection with text mining lie in the ambiguous semantics of natural languages and their various forms, character- istics and dynamics. Due to this it is difficult to formalize knowledge used in trend detection unambiguously and statically. Using ontologies, language com- ponents can be identified and subsequently processed and analyzed regarding their relations to each other. However, due to different languages and specific usages depending on user and application fields, as well as specific trend be- havior in certain application fields, trend ontologies specialized for the intended application are needed. In order to allow the modular development and usage of these different ontologies a standardizing base for trend ontologies is needed. This base can be realized as meta ontology and its development is the central aspect of the report.Weniger anzeigen
It is generally known that in the case of multiple node or communication failures atomic commit protocols cannot avoid blocking. While in wired networks such situations are rare because of low failure probabilities, mobile ad-hoch networks (MANETs) are considered to be a more challenging. In this technical report I present a probabilistic model to predict the abort and blocking risks of distrubted atomic transactions for arbitrary MANET scenarios. The model presented is applied to a standard MANET scenario to demonstrate the blocking risks to be expected.Weniger anzeigen
Computer simulations allow researchers to explore processes underlying neural functions. Biologically detailed simulations of single neurons and neural networks are based on compartmental modelling. This article describes seven well-known compartmental modelling packages and compares their characteristics. The important aspects of selecting a program are discussed.
Wireless mesh networks are an emerging and versatile communication technology. The most common application of these networks is to provide access of any number of users to the world wide Internet. They can be set up by Internet service providers or even individuals joined in communities. Due to the wireless medium that is shared by all participants, effects like short-time fading, or the multi-hop property of the network topology many issues are still in the focus of research. Testbeds are a powerful tool to study wireless mesh networks as close as possible to real world application scenarios. In this technical report we describe the design, architecture, and implementation of our work-in-progress wireless testbed at Freie Universität Berlin consisting of 100 mesh routers that span multiple buildings. The testbed is hybrid as it combines wireless mesh network routers with a wireless sensor network.Weniger anzeigen