Supervised sentiment analysis in Czech social media

This article describes in-depth research on machine learning methods for sentiment analysis of Czech social media. Whereas in English, Chinese, or Spanish this field has a long history and evaluation datasets for various domains are widely available, in the case of the Czech language no systematic research has yet been conducted. We tackle this issue and establish a common ground for further research by providing a large human-annotated Czech social media corpus. Furthermore, we evaluate state-of-the-art supervised machine learning methods for sentiment analysis. We explore different pre-processing techniques and employ various features and classifiers. We also experiment with five different feature selection algorithms and investigate the influence of named entity recognition and preprocessing on sentiment classification performance. Moreover, in addition to our newly created social media dataset, we also report results for other popular domains, such as movie and product reviews. We believe that this article will not only extend the current sentiment analysis research to another family of languages, but will also encourage competition, potentially leading to the production of high-end commercial solutions.     

Changes in joint kinetics during learning the longswing on high bar

Biomechanics helps us understand the association between technique changes and performance improvement during learning. The aim of this research was to investigate joint kinetic characteristics of technique during learning of the longswing on the high bar. Twelve male, novice participants took part in the learning study. During swing attempts in 8 weekly testing sessions, kinematic data were collected. Inverse dynamics analysis was performed from known zero forces at the toes to quantify joint moments and power at the hips and shoulders. Key biomechanical constraints that limited performance outcome were identified based on changes in joint kinetics during learning. These constraints were the ability to perform a large shoulder power and to overcome passive kinetics acting during the downswing. Constraints to action at the level of joint kinetics differentially challenge learners and therefore could underpin more individual, specific learning interventions. Functional phases, defined by maximum hyperextension to flexion of the hips and maximum flexion to extension of the shoulders, did not describe the key joint kinetics of the hip and shoulder for novices. The functional phases may serve however to identify novices that were unable to overcome the passive kinetic constraint.     

Biomechanical energetic analysis of technique during learning the longswing on the high bar

Biomechanical energetic analysis of technique can be performed to identify limits or constraints to performance outcome at the level of joint work, and to assess the mechanical efficiency of techniques. The aim of this study was to investigate the biomechanical energetic processes during learning the longswing on the high bar. Twelve male, novice participants took part in a training study. Kinematic and kinetics data were collected during swing attempts in eight weekly testing sessions. Inverse dynamics analysis was performed from known zero forces at the toes. Joint work, total energy, and bar energy were calculated. Biomechanical constraints to action, that is, limits to novice performance, were identified as “total work” and “shoulder work”. The most biomechanically efficient technique was associated with an onset of the hip functional phase and joint work that occurred between 10–45° before the bottom of the swing. The learning of gross motor skills is realised through the establishment of a set of techniques with task specific biomechanical constraints. Knowledge of the biomechanical constraints to action associated with more effective and efficient techniques will be useful for both assessing learning and establishing effective learning interventions.     

Digitization or equality: When government automation covers some,but not all citizens

This paper presents an empirical study of automation in government digital systems. Previous studies have found that automated systems are not suited to cover all citizens equally and may cause administrative burdens on excluded citizens. The case presented in this study is the automated system for awarding child benefits in Norway. Based on data from the national registry, most recipients are awarded the benefit automatically. However, some citizens are not covered by the automation and must apply manually. The theoretical framing of the study combines modern and classic views of how citizens access public services by combining theory from recent literature on administrative burdens and the older theory of access. The data analysis is done with process mining, an innovative method of sorting and understanding data. The findings support previous findings of how registry data and automated computer systems in government can create inequality in service quality. Furthermore, the findings also show that low-income citizens are disproportionally required to apply manually. The study addresses questions concerning why automated systems fail to cover all citizens and the potential challenges generated by this exclusion when governments rely on computer systems in delivering welfare programmes. These are important considerations, as government digitalisation is increasingly innovating with automated systems to deliver public services.     

Skill level changes the coordination and variability of standing posture and movement in a pistol-aiming task

The study investigated the coordination and variability of posture and pistol motion for skilled pistol shooters and novices in a pistol-aiming task. The participants stood on a force platform and held a pistol with the preferred arm to aim for accuracy to a target on 30 s trials. The results revealed that the amount of the centre of pressure (COP) and pistol motion was lower for the expert than novice group. The time–varying structure of COP as indexed by multiscale entropy (MSE) and detrended fluctuation analysis (DFA) was also lower for the expert than the novice group. The relative phase between the COP in the anterior–posterior (AP) and pistol in the AP and between the COP in the medial–lateral (ML) and pistol in AP was close to inphase for the both groups. However, for the novice group the coordination patterns of posture and pistol motion were more variable with the pistol motion leading the posture motion while it was lagging in the skilled group. The findings show different qualitative and quantitative dynamics in pistol-aiming as a function of skill level with postural control foundational to supporting the reduced dispersion and complexity of the skilled arm-pistol motion.     

L'institut technique communal pour l'industrie,le commerce et l'hotellerie de la ville d'ostende

Sans résumé