Posts by Collection

Several authorship analysis tasks require the decomposition of multi-authored text into its authorial components. Authorship identification is an important task within stylometry that can be applied to many cases. For example, determining the author of a ransom note can save someone’s life, discovering whether all the uploaded assignments of a student are classified as their own work can reduce the amount of plagiarism, but it can be also applied in arts to identify an author of an old text. The documents clustering task, by author’s linguistic style, is also of vital importance in forensic applications. In this project, we focus on unsupervised authorship analysis and provide an evaluation framework and a random baseline to compare different approaches. In this work, given a collection of short documents, we approach the author clustering task to determine which documents are written by the same author. The number of clusters is determined through the computation of silhouettes for some approaches. Several approaches are compared but Affinity Propagation clustering method has the best result with 0.51 average B-Cubed F-score without using n-gram features, and it is improved to 0.61 using n-gram features. Several features including Punctuations frequency, small tokens frequency, average tokens length, stop words frequency, Part of speech tags frequency and function words are extracted from data. Text data are gathered from 6 different Persian newspaper authors.

The main motivation for the extraordinary advances in the field of predictive process monitoring has been to enable data-driven prediction of potential violations in process performance and the subsequent improvement of processes in organizations through continuous process re-engineering in enterprises. The contribution of predictive process monitoring is instrumental and is a first step stone towards automating process performance improvement. Building on the advances in predictive process monitoring, in this paper we focus on the next steps towards autonomic process performance improvement. The other essential contribution towards this goal is served partly by the research in the field of process adaptation and flexibility. We show that the link and interplay between predictive monitoring and process adaptation, both in terms of underlying concepts and technological realization, have been inadequately researched, as revealed by our review of existing literature, leaving a huge gap towards completely automating the reaction to predictions produced by runtime monitoring techniques, or business experts for that matter. We also define and present a functional architecture that aims at assisting enterprises in developing solutions and the research communities when positioning their research findings in the broader fields of predictive process monitoring, process adaptation and enterprise architectures. Towards that goal, we also contribute a road map highlighting potential pitfalls and pointing to best engineering practices towards the realization of autonomic process performance improvement in enterprise systems.

Process-aware information systems are valuable for automating business tasks leading to cost reduction and efficiency. This research aims to advance the state of the art in process management towards autonomic process performance improvement by contributing control-flow change recommendations for process instances that is supporting automatic change enactment as a response to predicted KPI violations. Towards that goal, the related literature has been investigated in two literature review studies and research gaps have been identified. The proposed generic architecture provides a feedback loop that enables evaluation of the resulting recommendations for future process instances. We also present the current state of the research and future plans.

In this paper we present a tool for adaptive process log generation and analysis of the correlation between KPI (Key Performance Indicator) values and changes in adaptive processes. The tool features a component called Next(Log) helping users to generate initial business process logs using any preferred method and subsequently allows them to adapt these logs based on their own defined rules while ensuring an intuitive and coherent user interface. The adapted logs are then used for log analysis with the ML.Log component, which employs machine learning techniques to find patterns of matching KPI values and adaptation injections in the logs. The tool therefore supports the research on the challenges imposed by the lack of sufficient amount of data from adaptive process logs and the open issues in identifying at what KPIs values changes are required and what kind of changes would have the best impact on the process performance at run time.