Carter, A.; Nasir, W.; Parker, E. Domain Adaptation with Sentiment Domain Adapter. Preprints2024, 2024041031. https://doi.org/10.20944/preprints202404.1031.v1
APA Style
Carter, A., Nasir, W., & Parker, E. (2024). Domain Adaptation with Sentiment Domain Adapter. Preprints. https://doi.org/10.20944/preprints202404.1031.v1
Chicago/Turabian Style
Carter, A., Wyne Nasir and Ethan Parker. 2024 "Domain Adaptation with Sentiment Domain Adapter" Preprints. https://doi.org/10.20944/preprints202404.1031.v1
Abstract
The field of domain adaptation, particularly in cross-domain sentiment classification, leverages labeled data from a source domain alongside unlabeled or sparsely labeled data from a target domain. The objective is to enhance performance in the target domain by mitigating the discrepancies in data distributions. Traditional methods in this area have focused on identifying and differentiating between pivotal sentiment words (shared across domains) and domain-specific sentiment words. In our work, we introduce a novel framework called Sentiment Domain Adapter (SDA), which incorporates a Category Attention Network (CAN) alongside a Convolutional Neural Network (CNN). This approach treats pivotal and domain-specific words as part of a collective category of attributes, which SDA learns to discern automatically, thereby enhancing domain adaptation. Additionally, SDA seeks to provide interpretative insights by learning these category attributes. Our model's optimization targets three main goals: (1) supervised classification accuracy, (2) minimizing the disparity in category feature distribution, and (3) maintaining domain invariance. Evaluations on three benchmark datasets for sentiment analysis affirm that SDA surpasses several established baselines.
Computer Science and Mathematics, Artificial Intelligence and Machine Learning
Copyright:
This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
