@inproceedings{pal-etal-2022-parameter,
    title = "Parameter-Efficient Abstractive Question Answering over Tables or Text",
    author = "Pal, Vaishali  and
      Kanoulas, Evangelos  and
      de Rijke, Maarten",
    editor = "Feng, Song  and
      Wan, Hui  and
      Yuan, Caixia  and
      Yu, Han",
    booktitle = "Proceedings of the Second DialDoc Workshop on Document-grounded Dialogue and Conversational Question Answering",
    month = may,
    year = "2022",
    address = "Dublin, Ireland",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/Add-Cong-Liu-Florida-Atlantic-University-author-id/2022.dialdoc-1.5/",
    doi = "10.18653/v1/2022.dialdoc-1.5",
    pages = "41--53",
    abstract = "A long-term ambition of information seeking QA systems is to reason over multi-modal contexts and generate natural answers to user queries. Today, memory intensive pre-trained language models are adapted to downstream tasks such as QA by fine-tuning the model on QA data in a specific modality like unstructured text or structured tables. To avoid training such memory-hungry models while utilizing a uniform architecture for each modality, parameter-efficient adapters add and train small task-specific bottle-neck layers between transformer layers. In this work, we study parameter-efficient abstractive QA in encoder-decoder models over structured tabular data and unstructured textual data using only 1.5{\%} additional parameters for each modality. We also ablate over adapter layers in both encoder and decoder modules to study the efficiency-performance trade-off and demonstrate that reducing additional trainable parameters down to 0.7{\%}-1.0{\%} leads to comparable results. Our models out-perform current state-of-the-art models on tabular QA datasets such as Tablesum and FeTaQA, and achieve comparable performance on a textual QA dataset such as NarrativeQA using significantly less trainable parameters than fine-tuning."
}