@inproceedings{nadiger-etal-2026-momentum,
    title = "Momentum at {S}em{E}val-2026 Task 2: {L}ong{VA}-{R}o{BERT}a, a transformer-Based Longitudinal Valence and Arousal Modeling",
    author = "Nadiger, Supriya  and
      Saumya, Sunil  and
      Pujari, Rahul  and
      Hiremath, Veeresh  and
      Chikaraddi, Kiran  and
      Kadkol, Anoop",
    editor = "Kochmar, Ekaterina  and
      Ghosh, Debanjan  and
      North, Kai  and
      Komachi, Mamoru",
    booktitle = "Proceedings of the 20th {I}nternational {W}orkshop on {S}emantic {E}valuation (2026)",
    month = jul,
    year = "2026",
    address = "San Diego, California, USA",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/ingest-acl-workshops/2026.semeval-1.78/",
    pages = "546--552",
    ISBN = "979-8-89176-414-9",
    abstract = "This paper studies the emotion as affective circumplex model representing valence and arousal in continuous two dimensional space. It also explores the disposition of emotion over time to identify the behavioural cues and self-identified affective states. while traditional methods use categorical emotion classes, SemEval 2026 Task 2 studies emotions in continuous space. In this paper, we proposes a transformer-based LongVA-RoBERTa model for emotion modeling in regression for ecological essays. For subtask 1 , we develop an affect prediction framework employing RoBERTa with attention pooling and a regression head for valence and arousal prediction. In subtask 2A , we employ BiLSTM to capture the temporal dependencies and fuse surface, contextual, user-level features to predict short-term affect variation. Our results outperform the baseline, paving ways to continue emotion prediction in continuous dimensional space"
}