Keeping in Time: Adding Temporal Context to Sentiment Analysis Models

TL;DR

This paper introduces a temporal context-aware sentiment analysis framework that incorporates date information into language models, improving performance across different time periods and demonstrating state-of-the-art results in the LongEval-Classification task.

Contribution

It presents a novel date-prefixed input method combined with self-labeling and augmentation strategies to enhance temporal robustness in sentiment analysis models.

Findings

Achieved 2nd place in LongEval-Classification with a score of 0.6923.

Demonstrated improved performance over non-augmented self-labeling.

Reported the best Relative Performance Drop (RPD) of -0.0656.

Abstract

This paper presents a state-of-the-art solution to the LongEval CLEF 2023 Lab Task 2: LongEval-Classification. The goal of this task is to improve and preserve the performance of sentiment analysis models across shorter and longer time periods. Our framework feeds date-prefixed textual inputs to a pre-trained language model, where the timestamp is included in the text. We show date-prefixed samples better conditions model outputs on the temporal context of the respective texts. Moreover, we further boost performance by performing self-labeling on unlabeled data to train a student model. We augment the self-labeling process using a novel augmentation strategy leveraging the date-prefixed formatting of our samples. We demonstrate concrete performance gains on the LongEval-Classification evaluation set over non-augmented self-labeling. Our framework achieves 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.