中文

文档自动化架构:基于大语言模型的更新综述

计算与语言 2023-08-21 v1 机器学习

摘要

本文综述了文档自动化(DA)领域的最新技术水平。DA 的目标是通过自动创建并整合来自不同来源的输入、以及按照既定模板组装文档,从而减少文档生成过程中的人工工作量。已有关于 DA 商业解决方案的综述,尤其在法律领域,但迄今为止尚无关于 DA 架构与技术的学术研究的综合性综述。本 DA 综述回顾了学术文献,给出了 DA 及其特征的更清晰定义与刻画,识别了学术研究中最先进的 DA 架构与技术,并基于生成式 AI 与大语言模型(LLM)的最新进展,提出了可在 DA 领域内催生新研究机遇的思路。

关键词

引用

@article{arxiv.2308.09341,
  title  = {Document Automation Architectures: Updated Survey in Light of Large Language Models},
  author = {Mohammad Ahmadi Achachlouei and Omkar Patil and Tarun Joshi and Vijayan N. Nair},
  journal= {arXiv preprint arXiv:2308.09341},
  year   = {2023}
}

备注

The current paper is the updated version of an earlier survey on document automation [Ahmadi Achachlouei et al. 2021]. Updates in the current paper are as follows: We shortened almost all sections to reduce the size of the main paper (without references) from 28 pages to 10 pages, added a review of selected papers on large language models, removed certain sections and most of diagrams. arXiv admin note: substantial text overlap with arXiv:2109.11603