Large, Small or Both: A Novel Data Augmentation Framework Based on Language Models for Debiasing Opinion Summarization

arXiv:2403.07693v1 Announce Type: new
Abstract: As more than 70$\%$ of reviews in the existing opinion summary data set are positive, current opinion summarization approaches are reluctant to generate negative summaries given the input of negative texts. To address such sentiment bias, a direct approach without the over-reliance on a specific framework is to generate additional data based on large language models to balance the emotional distribution of the dataset. However, data augmentation based on large language models faces two disadvantages: …

abstract arxiv augmentation bias cs.ai cs.cl current data data set framework generate language language models negative novel opinion positive reviews sentiment set small summarization summary type