The paper presents a state-of-the-art solution for the LongEval CLEF 2023 Lab Task 2. The goal is to enhance the performance of sentiment analysis models over time by adding temporal context. The approach involves feeding date-prefixed textual inputs to a pre-trained language model. This method conditions model outputs based on the temporal context of the texts. Performance is further improved by performing self-labeling on unlabeled data to train a student model. A novel augmentation strategy is used, leveraging the date-prefixed formatting of samples. The framework achieved a 2nd place ranking with an overall score of 0.6923 and reports the best Relative Performance Drop (RPD) of -0.0656 over the short evaluation set.
Publication date: 24 Sep 2023
Project Page: http://ceur-ws.org
Paper: https://arxiv.org/pdf/2309.13562
