# Quantifying Valence and Arousal in Text with Multilingual Pre-trained Transformers

arXiv: https://arxiv.org/abs/2302.14021 · HTML: https://mlsanity.com/paper/2302.14021/

## Key facts

- Cited by 11 papers in the map (data through 7 Oct 2026).
- Builds on 10 of its 68 references that are in the map.
- First posted on arXiv 27 Feb 2023.

## Builds on (most cited first, 10 of 10)

- [DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter](https://mlsanity.com/paper/1910.01108/) (2019)
- [Unsupervised Cross-lingual Representation Learning at Scale](https://mlsanity.com/paper/1911.02116/) (2019)
- [A Gentle Introduction to Conformal Prediction and Distribution-Free Uncertainty Quantification](https://mlsanity.com/paper/2107.07511/) (2021)
- [How multilingual is Multilingual BERT?](https://mlsanity.com/paper/1906.01502/) (2019)
- [A General and Adaptive Robust Loss Function](https://mlsanity.com/paper/1701.03077/) (2017)
- [EmoBank: Studying the Impact of Annotation Perspective and Representation Format on Dimensional Emotion Analysis](https://mlsanity.com/paper/2205.01996/) (2022)
- [Building Emotional Machines: Recognizing Image Emotions through Deep Neural Networks](https://mlsanity.com/paper/1705.07543/) (2017)
- [Evaluation of Error and Correlation-Based Loss Functions For Multitask Learning Dimensional Speech Emotion Recognition](https://mlsanity.com/paper/2003.10724/) (2020)
- [Learning and Evaluating Emotion Lexicons for 91 Languages](https://mlsanity.com/paper/2005.05672/) (2020)
- [A Multi-task Ensemble Framework for Emotion, Sentiment and Intensity Prediction](https://mlsanity.com/paper/1808.01216/) (2018)

## Built on by (most cited first, 11 of 11)

- [Fine-grained Affective Processing Capabilities Emerging from Large Language Models](https://mlsanity.com/paper/2309.01664/) (2023)
- [Valence-Arousal Subspace in LLMs: Circular Emotion Geometry and Multi-Behavioral Control](https://mlsanity.com/paper/2604.03147/) (2026)
- [Don't Get Too Excited -- Eliciting Emotions in LLMs](https://mlsanity.com/paper/2503.02457/) (2025)
- [A Computational Analysis of the Dehumanisation of Migrants from Syria and Ukraine in Slovene News Media](https://mlsanity.com/paper/2404.07036/) (2024)
- [Emotion Embeddings unicodex2014 Learning Stable and Homogeneous Abstractions from Heterogeneous Affective Datasets](https://mlsanity.com/paper/2308.07871/) (2023)
- [Posts of Peril: Detecting Information About Hazards in Text](https://mlsanity.com/paper/2405.17838/) (2024)
- [Mind the Style: Impact of Communication Style on Human-Chatbot Interaction](https://mlsanity.com/paper/2602.17850/) (2026)
- [A Hormone-inspired Emotion Layer for Transformer language models (HELT)](https://mlsanity.com/paper/2605.13858/) (2026)
- [From Affect Prediction to Affect Forecasting: Evidence for Distinct Information Sources in Longitudinal Text](https://mlsanity.com/paper/2606.16687/) (2026)
- [VIBE: A VAD-Informed Benchmark for Entity-Centered Affective Profiling of Large Language Model Outputs](https://mlsanity.com/paper/2608.03810/) (2026)
- [EmoLASP: Emotion Recognition with Language Models and Answer Set Programming](https://mlsanity.com/paper/2608.29035/) (2026)

## Abstract (arXiv)

The analysis of emotions expressed in text has numerous applications. In contrast to categorical analysis, focused on classifying emotions according to a pre-defined set of common classes, dimensional approaches can offer a more nuanced way to distinguish between different emotions. Still, dimensional methods have been less studied in the literature. Considering a valence-arousal dimensional space, this work assesses the use of pre-trained Transformers to predict these two dimensions on a continuous scale, with input texts from multiple languages and domains. We specifically combined multiple annotated datasets from previous studies, corresponding to either emotional lexica or short text documents, and evaluated models of multiple sizes and trained under different settings. Our results show that model size can have a significant impact on the quality of predictions, and that by fine-tuning a large model we can confidently predict valence and arousal in multiple languages. We make available the code, models, and supporting data.