Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

De positie van de toeschouwer is binnen het academische discours telkens weer anders geconcipieerd. Vergeleken met andere aspecten van het theaterwezen is het publiek grotendeels over het hoofd gezien binnen de theaterwetenschappen. Daar is pas recentelijk verandering in gebracht. Deze verzameling essays werpt een licht op hedendaagse ontwikkelingen binnen de filosofie, theater en performance studies met betrekking tot de positie van de toeschouwer.

Arts du spectacle --- Performing arts --- Podiumkunsten --- Theatrical science --- Spectators. --- Audiences. --- Communication --- Management

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

The problem at hand is presented by the Youth Skills research team, a European funded project that aims to 'enhance and maximise long-term positive impact of the digital environment'. They have presented the results of a survey in which young teenagers have assessed the accuracy, authenticity and the credibility of online news articles on a scale from 1 to 7. Along with the numerical assessment, the students were asked to motivate the numbers in a written comment. The resulting data set is small and the answers are concise.ewline egin{comment} Based on this data, the Youth Skills researchers would like to create a 'Youth Skills score' that gives feedback on the digital literacy experience of individual students. The first step in the machine-aided establishment of such a score, is the development of a mathematical model that can recognize to which degree a student assesses online information. From a machine learning perspective, the problem at hand is known as short-text classification and has gained increasing interest in the community over the past years. Although the focus of the community is mainly aimed at big data, contributions have been made in the form of case studies that train models on moderate and small data sets. This is especially the case in the context of automated grading in education. end{comment} Models trained on small sized data sets with short answers, suffer from feature instability and hence bad classification performance if they are not constructed with care. The final performance does not only depend on the type of machine learning model. Also text representation, feature selection and feature selection stability are significantly important to this specific problem setting. Each of these aspects in the training pipeline are accounted for. In total, the Youth Skills problem is evaluated using the following eight types of classifiers : Support Vector Classifier (SVC), Support Vector Regressor (SVR), Multinomial Naive Bayes (MNB), Gaussian Naive Bayes (GNB), Random Forrest Classifier (RFC), Random Forrest Regressor (RFR), Multi-Layer Perceptron Classifier (MLPC) and Multi-Layer Perceptron Regressor (MLPR). Each model is combined with two feature selection algorithms, Document Frequency (DF) and extChiextsuperscript{2} (CHI) on three similar data sets ('ACC', 'AUTH' and 'BEL'). Text representation is established with the Bag-of-Words (BOW) model in combination with Text Frequency-Inverse Document Frequency (TF-IDF) vectorization. Finally, feature selection stability is accounted for with an inner K-Fold implementation and similarity calculations using the Jaccard index. Unfortunately, the results did not fulfill our expectations. Based on the accuracy, none of the resulting models is considered viable for practical implementation in the ySKILLS research. The underlying causes are open to speculation and we address some of them. Nevertheless, reasonable feature selection stability is observed in a significant number of cases. The stable features may be of use to the ySKILLS project as discussed towards the end of this paper. Finally, several options for future improvements are added.