As part of the institution’s 2024–2027 strategic plan, which includes the objective of understanding how the media portrays the organization to strengthen its public image, this paper investigates the application of deep learning algorithms in sentiment analysis of headline news about a public security institution. Four deep learning methods were applied in combination with three textual representations, resulting in twelve trained models. For each combination, a class-based analysis of the results was conducted. Models using BERT as the textual representation achieved strong performance, with an F1-score of approximately 90%.
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。