Dissertation: Einsatz künstlicher neuronaler Netze zur Identifikation von Ironie im Rahmen der Sentiment-Analyse

Einsatz künstlicher neuronaler Netze zur Identifikation von Ironie im Rahmen der Sentiment-Analyse

Studien zur Wirtschaftsinformatik, Band 105

Hamburg 2020, 312 Seiten
ISBN 978-3-339-11760-1 (Print), ISBN 978-3-339-11761-8 (eBook)

Ironie, Ironieforschung, Künstliche Intelligenz, Künstliche neuronale Netze, Maschinelles Lernen, Sarkasmus, Sentiment-Analyse, Textanalyse

about this book

deutsch | english

The automated identification of irony poses a complex challenge for computer-aided systems, since as a figurative speech act it requires extensive cognitive skills for recognition. At the same time, the correct detection of irony is of great importance in various areas of application of computer-aided data processing and analysis, especially in the area of sentiment analysis, since disregarding the presence of irony can lead to considerable falsifications of analysis results. Artificial neural networks, as a replica of natural neural structures, enable computer-aided systems to solve problems that require complex cognitive skills, including the detection of irony.

The book at hand deals with the phenomenon of irony and its automated computer-aided identification using artificial neural networks. Based on the findings of an analysis of previous research work, there is a structured development of an artificial neural network, which is able to identify irony in text data. The development process includes all essential aspects that influence the performance of an artificial neural network, including the basic structure of the network, the choice of hyperparameters, the use of regulation techniques and data preprocessing including the design of the features, which represent die input data to the network. The development process is followed by a test phase that evaluates the developed artificial neural network on the basis of different datasets with ironic content. In the course of the test phase, the existing pool of ironic datasets was expanded by a new dataset, which for the first time contains German-language ironic text data.

Informationen über das Veröffentlichen wissenschaftlicher Arbeiten.

nach oben