Improving sentiment analysis with electroencephalography (EEG) features / Fall 2018 - Master Thesis
It has been shown that various Natural Language Processing (NLP) tasks can benefit from cognitive features. We have recorded a new dataset  of simultaneous eye-tracking and EEG recording while subjects are reading natural sentences occurring in movie reviews from the Stanford Sentiment Treebank .
Now we want to build a neural sentiment analysis system that benefits from these EEG features. While we have initial results from a previous master thesis which show great potential, EEG signals have not been explored in detail for machine learning applications. Thus, this work includes not only building a sentiment analysis system but also a feature extraction and analysis study for EEG data. Furthermore, an additional goal of this project is to figure out how to use the EEG signal only at training data, without needing EEG signals for test sentences.
Supervised by Nora Hollenstein
-  Hollenstein, N., Langer, N., Pedroni, A., Troendle, M., Rotsztejn, J., & Zhang, C. (2018, May 28). Zurich Cognitive Language Processing Corpus: A simultaneous EEG and eye-tracking resource to analyze the human reading process. Retrieved from osf.io/q3zws
-  Socher, R., Perelygin, A., Wu, J., Chuang, J., Manning, C. D., Ng, A., & Potts, C. (2013). Recursive deep models for semantic compositionality over a sentiment treebank. In Proceedings of the 2013 conference on empirical methods in natural language processing (pp. 1631-1642).