Modelling Representation Noise in Emotion Analysis using Gaussian Processes
Daniel Beck · International Joint Conference on Natural Language Processing · 2017
Emotion Analysis is the task of modelling la-tent emotions present in natural language. La-belled datasets for this task are scarce so learn-ing good input text representations is not triv-ial. Using averaged word embeddings is a sim-ple way to leverage unlabelled corpora to build text representations but this approach can be prone to noise either coming from the embed-ding themselves or the averaging procedure. In this paper we propose a model for Emotion Analysis using Gaussian Processes and kernels that are better suitable for functions that ex-hibit noisy behaviour. Empirical evaluations in a emotion prediction task show that our model outperforms commonly used baselines for re-gression.