@InProceedings{he-shi:2026:NeTTIT,
  author    = {HE, Sihui  and  SHI, Jiawei},
  title     = {Exploring Sentiment Transfer in Cantonese–English Court Interpreting: A Computational Analysis Using Multilingual BERT},
  booktitle      = {Proceedings of the Conference on New Trends in Translation and Interpreting Technology 2026},
  month          = {June},
  year           = {2026},
  address        = {Dubrovnik, Croatia},
  publisher      = {INCOMA Ltd., Shoumen, Bulgaria},
  pages     = {93--97},
  abstract  = {The way sentiment is conveyed during court interpreting may result in different legal outcomes. This study used multilingual BERT to analyze the sentiment of the Hong Kong Court Interpreting Corpus (HKCI). The corpus comprises 101 audio recordings from five rape trials at the High Court of Hong Kong. The results showed that negative sentiment predominated across all four sub-corpora and all pairwise comparisons were statistically significant. Moreover, in the C2E direction, the negative sentiment was reduced, while in the E2C direction, increased negative sentiment was observed. This pattern may be partly explained by the structural differences between Cantonese and English. In addition, institutional power status was not found to predict sentiment shift in either direction. Furthermore, BERT predictions showed low agreement with human annotations (κ = 0.19). These findings may provide insights for court interpreting. The intensified negative sentiment in E2C output suggests witnesses and defendants may perceive more negative renditions of legal professionals' speech, which raises potential concerns for procedural justice. These findings also indicate the need to develop domain-adapted sentiment tools for low-resource legal discourse.},
  url       = {https://aclanthology.org/2026.nettit-1.13}
}