# Deep Bag-of-Sub-Emotions for Depression Detection in Social Media

arXiv: https://arxiv.org/abs/2103.01334 · HTML: https://mlsanity.com/paper/2103.01334/

## Key facts

- Cited by 4 papers in the map (data through 14 Oct 2026).
- Builds on 5 of its 25 references that are in the map.
- First posted on arXiv 1 Mar 2021.

## Builds on (most cited first, 5 of 5)

- [Visualizing and Understanding Neural Models in NLP](https://mlsanity.com/paper/1506.01066/) (2015)
- [Crowdsourcing a Word-Emotion Association Lexicon](https://mlsanity.com/paper/1308.6297/) (2013)
- [Forecasting the onset and course of mental illness with Twitter data](https://mlsanity.com/paper/1608.07740/) (2016)
- [Semi-Supervised Approach to Monitoring Clinical Depressive Symptoms in Social Media](https://mlsanity.com/paper/1710.05429/) (2017)
- [Dissimilarity Mixture Autoencoder for Deep Clustering](https://mlsanity.com/paper/2006.08177/) (2020)

## Built on by (most cited first, 4 of 4)

- [It's Just a Matter of Time: Detecting Depression with Time-Enriched Multimodal Transformers](https://mlsanity.com/paper/2301.05453/) (2023)
- [Emotion fusion for mental illness detection from social media: A survey](https://mlsanity.com/paper/2304.09493/) (2023)
- [Life is not Always Depressing: Exploring the Happy Moments of People Diagnosed with Depression](https://mlsanity.com/paper/2204.13569/) (2022)
- [3M-Health: Multimodal Multi-Teacher Knowledge Distillation for Mental Health Detection](https://mlsanity.com/paper/2407.09020/) (2024)

## Abstract (arXiv)

This paper presents the Deep Bag-of-Sub-Emotions (DeepBoSE), a novel deep learning model for depression detection in social media. The model is formulated such that it internally computes a differentiable Bag-of-Features (BoF) representation that incorporates emotional information. This is achieved by a reinterpretation of classical weighting schemes like term frequency-inverse document frequency into probabilistic deep learning operations. An important advantage of the proposed method is that it can be trained under the transfer learning paradigm, which is useful to enhance conventional BoF models that cannot be directly integrated into deep learning architectures. Experiments were performed in the eRisk17 and eRisk18 datasets for the depression detection task; results show that DeepBoSE outperforms conventional BoF representations and it is competitive with the state of the art, achieving a F1-score over the positive class of 0.64 in eRisk17 and 0.65 in eRisk18.