|
|
Social Network Analysis using methods of pattern recognition
Križan, Viliam ; Burget, Radim (oponent) ; Atassi, Hicham (vedoucí práce)
The diploma thesis deals with emotion recognition from texts on social media. The state-of-the-art methods of feature extraction, corpora and classifiers are described in the first section. Emotions are recognized by a classifier trained on annotated data from the microblog network Twitter. The advantage of using Twitter was the possibility to specify data collection to a certain geographical location. Geographical data allows to monitor emotional variations of population, for e.g. in different cities. The first task was to propose and develop a Baseline algorithm which classifies data to emotional classes. The classification accuracy is improved by employing a more complex SVM classifier. SVM classifiers, feature vectorizers and feature selectors are used from the Scikit library, which is written in Python. The data for classifier training were collected from the USA by the own developed mining application. The classifier are trained on data automatically annotated in the collection process. Two implementations of SVM classifiers are used. Final classified emotions that appear in different cities and in different time intervals are displayed as color markers on a map.
|
|
|
Nové metody zpracování textu pro klasifikaci emocí
Onderka, Jakub ; Burget, Radim (oponent) ; Mašek, Jan (vedoucí práce)
Tato diplomová práce se zabývá možnostmi extrakce emocí z textu, konkrétně strojovými metodami bez učitele. Podrobněji jsou popsány metody sémantického modelování LSA, pLSA a LDA. Byla vytvořena implementace metody LDA v jazyce Java, která byla použita pro emocionální klasifikaci 860 česky psaných dokumentů do šesti odlišných emocí. Maximální přesnost při optimalizaci parametrů modelu byla 24 %.
|
|
|
Nové metody zpracování textu pro klasifikaci emocí
Onderka, Jakub ; Burget, Radim (oponent) ; Mašek, Jan (vedoucí práce)
Tato diplomová práce se zabývá možnostmi extrakce emocí z textu, konkrétně strojovými metodami bez učitele. Podrobněji jsou popsány metody sémantického modelování LSA, pLSA a LDA. Byla vytvořena implementace metody LDA v jazyce Java, která byla použita pro emocionální klasifikaci 860 česky psaných dokumentů do šesti odlišných emocí. Maximální přesnost při optimalizaci parametrů modelu byla 24 %.
|
|
|
Social Network Analysis using methods of pattern recognition
Križan, Viliam ; Burget, Radim (oponent) ; Atassi, Hicham (vedoucí práce)
The diploma thesis deals with emotion recognition from texts on social media. The state-of-the-art methods of feature extraction, corpora and classifiers are described in the first section. Emotions are recognized by a classifier trained on annotated data from the microblog network Twitter. The advantage of using Twitter was the possibility to specify data collection to a certain geographical location. Geographical data allows to monitor emotional variations of population, for e.g. in different cities. The first task was to propose and develop a Baseline algorithm which classifies data to emotional classes. The classification accuracy is improved by employing a more complex SVM classifier. SVM classifiers, feature vectorizers and feature selectors are used from the Scikit library, which is written in Python. The data for classifier training were collected from the USA by the own developed mining application. The classifier are trained on data automatically annotated in the collection process. Two implementations of SVM classifiers are used. Final classified emotions that appear in different cities and in different time intervals are displayed as color markers on a map.
|
host ::
RSS.