This paper aims to introduce the emerging artificial-intelligence-based readability metrics (Coh-Metrix) to examine the effects of firm size on audit proposal readability.,Coh-Metrix readability measures use emerging computation linguistics technology to better assess document readability. These metrics measure co-relations of words, sentences and paragraphs on multi-dimensions rather than adopting the unidimensional “bag of words” approach that examines words in isolation. Using eight Coh-Metrix orthogonal principal component factors, the authors analyze the Chang and Stone (2019) data set comprised of 370 hand-collected audit proposals submitted by audit firms for the US state and local governments’ audit service contracts.,Audit firm size has a significant impact on the readability of audit proposals. Specifically, as measured by the traditional readability metric, the proposals from smaller firms are more readable than those submitted by larger firms. Furthermore, decomposed readability metrics indicate that smaller firm proposals evidence stronger (deep) text cohesion, whereas larger firm proposals evidence a stronger narrative structure and higher connectivity (relational indicators) among proposal elements. Unlike the traditional readability metric, however, the emergent readability metrics are uncorrelated with auditor selection.,Work remains to develop and validate Coh-Metrix measures that are specific to the context of accounting and auditing practice. Future research can use emerging readability measures to examine various textual features (e.g. text cohesion) in finance or accounting related documents.,The results provide practitioners with insight into the proposal writing strategies and practices of larger and smaller firms. In addition, the results highlight the differing audit firm selection outcomes from traditional and Coh-Metrix readability metrics.,This study introduces new data and holistic readability measures to the auditing literature.

中文翻译：

---

为什么分解后的审计建议可读性随审计公司规模的不同而不同？Coh-Metrix方法

本文旨在介绍新兴的基于人工智能的可读性度量标准（Coh-Metrix），以研究公司规模对审计建议可读性的影响。Coh-Metrix可读性度量使用新兴的计算语言技术来更好地评估文档的可读性。这些度量标准衡量单词，句子和段落在多维上的相互关系，而不是采用一维的“单词袋”方法来孤立地检查单词。作者使用八个Coh-Metrix正交主成分因子，对Chang和Stone（2019）数据集进行了分析，该数据集由370家由审计公司针对美国各州和地方政府提供的审计服务合同手工收集的审计建议组成。对审计建议的可读性有重大影响。特别，按照传统的可读性度量标准，小公司的建议比大公司的建议更具可读性。此外，分解后的可读性度量标准表明，较小的公司建议书显示较强的（深度）文本内聚力，而较大的公司建议书则显示较强的叙述结构和建议元素之间较高的连通性（关系指标）。但是，与传统的可读性度量标准不同，新兴的可读性度量标准与审计师的选择无关。仍然需要开发和验证针对会计和审计实践背景的Coh-Metrix度量。未来的研究可以使用新兴的可读性度量来检查与财务或会计相关的文档中的各种文本特征（例如，文本内聚性）。结果为从业人员提供了对规模较大的公司的提案撰写策略和实践的洞察力。此外，结果突出了与传统的和Coh-Metrix可读性指标不同的审计公司选择结果。本研究向审计文献介绍了新的数据和整体可读性度量。