Abstract
We build and analyze a coreference network based on entities from photo descriptions, where nodes represent personalities and edges connect people mentioned in the same photo description. We identify and characterize the communities in this network and propose taking advantage of the context provided by community detection methodologies to improve text illustration and general search. © 2012 Springer-Verlag Berlin Heidelberg.

Abstract
Text illustration is a multimedia retrieval task that consists in finding suitable images to illustrate text fragments such as blog entries, news reports or children stories. In this paper we describe a crossmedia retrieval system which, given a textual input, selects a short list of candidate images from a large media collection. This approach makes use of a recently proposed method to map metadata and visual features into a common textual representation that can be handled by traditional information retrieval engines. Content-based analysis is enhanced by visual abstraction, namely the Anisotropic Kuwahara Filter, which impacts feature information captured by the Joint Composite and Speeded Up Robust Features visual descriptors. For evaluation purposes, we used the well-established MIRFlickr photo collection, with 25,000 photos and user tags collected from Flickr as well as manual annotations provided as image retrieval groundtruth. Results show that image abstraction can improve visual retrieval as well as significantly reduce processing and storage requirements, even more when paired with Google's WebP image format. We conclude that applying a visual rerank after an initial text retrieval step improves the quality of results, and that the adopted text mapping method for visual descriptors provides an effective crossmedia approach for text illustration. © 2012 Springer-Verlag Berlin Heidelberg.

Abstract

Abstract
High-maturity software development processes, making intensive use of metrics and quantitative methods, such as the Personal Software Process (PSP) and the Team Software Process (TSP), can generate a significant amount of data that can be periodically analyzed to identify performance problems, determine their root causes and devise improvement actions. Currently, there are several tools that automate data collection and produce performance charts for manual analysis in the context of the PSP/TSP, but practically no tool support exists for automating the data analysis and the recommendation of improvement actions. Manual analysis of this performance data is problematic because of the large amount of data to analyze and the time and expertise required. Hence, we propose in this paper a performance model and a tool (named PSP PAIR) to automate the analysis of performance data produced in the context of the PSP, namely, identify performance problems and their root causes, and recommend improvement actions. The work presented is limited to the analysis of the time estimation performance of PSP developers, but is extensible to other performance indicators and development processes.

Abstract
This paper proposes a new approach to reduce the effort of building formal models representative of the structure and behaviour of Graphical User Interfaces (GUI). The main goal is to automatically extract the GUI model with a dynamic reverse engineering process, consisting in an exploration phase, that extracts information by interacting with the GUI, and in a model generation phase that, making use of machine learning techniques, uses the extracted information of the first step to generate a state-machine model of the GUI, including guard conditions to remove ambiguity in transitions. © 2012 IEEE.

Abstract
Model-driven engineering approaches aim at avoiding productivity, model quality and model maintenance problems that arise when models are used for documentation only, by generating executable applications from models. However, in many cases, the level of detail of the models needed to generate complete applications is too much or only effective for specific domains. For those cases where it is not practical to build complete models and generate complete applications from them, we propose a lightweight approach, applicable at different levels (unit, integration and system testing), that combines partial application generation from structural models with test generation from partial behavioral models. To demonstrate the approach, we developed a plug-in that adds to the code generation capabilities of an existing UML modeling tool, the capability of generating executable tests from sequence diagrams acting also as parameterized test scenarios, including some novel features as compared to existing model-based testing tools.