Towards Label-Agnostic Emotion Embeddings
This paper addresses the problem of data and software heterogeneity in emotion analysis, making it easier for researchers to compare and integrate resources across different label formats, languages, and models.
The authors propose a training scheme to learn a shared latent representation of emotion that is independent of label formats, natural languages, and model architectures. Experiments show this approach achieves interoperability without sacrificing prediction quality.
Research in emotion analysis is scattered across different label formats (e.g., polarity types, basic emotion categories, and affective dimensions), linguistic levels (word vs. sentence vs. discourse), and, of course, (few well-resourced but much more under-resourced) natural languages and text genres (e.g., product reviews, tweets, news). The resulting heterogeneity makes data and software developed under these conflicting constraints hard to compare and challenging to integrate. To resolve this unsatisfactory state of affairs we here propose a training scheme that learns a shared latent representation of emotion independent from different label formats, natural languages, and even disparate model architectures. Experiments on a wide range of datasets indicate that this approach yields the desired interoperability without penalizing prediction quality. Code and data are archived under DOI 10.5281/zenodo.5466068.