Textual Analysis in Accounting: What's Next?*

Abstract

Natural language is a key form of business communication. Textual analysis is the application of natural language processing (NLP) to textual data for automated information extraction or measurement. We survey publications in top accounting journals and describe the trend and current state of textual analysis in accounting. We organize available NLP methods in a unified framework. Accounting researchers have often used textual analysis to measure disclosure sentiment, readability, and disclosure quantity; to compare disclosures to determine similarities or differences; to identify forward‐looking information; and to detect themes. For each of these tasks, we explain the conventional approach and newer approaches, which are based on machine learning, especially deep learning. We discuss how to establish the construct validity of text‐based measures and the typical decisions researchers face in implementing NLP models. Finally, we discuss opportunities for future research. We conclude that (i) textual analysis has grown as an important research method and (ii) accounting researchers should increase their knowledge and use of machine learning, especially deep learning, for textual analysis.

FinBERT: A Large Language Model for Extracting Information from Financial Text*

“…We choose these texts for three reasons. First, they are the most representative and frequently used types of corporate financial communications in financial economics research (Li 2010b; Loughran and McDonald 2016; Bochkay et al 2023). Including them in pretraining helps FinBERT analyze textual information most relevant to research in these areas.…”

“…A burgeoning finance and accounting literature has explored the use of natural language processing (NLP) algorithms to analyze financial texts (see reviews by Li 2010b; Das 2014; Loughran and McDonald 2016; Gentzkow et al 2019; Bochkay et al 2023). Most, if not all, of these studies rely on NLP algorithms that assume a bag‐of‐words structure 1 .…”

“…This study primarily contributes to the growing body of literature that uses textual analysis in financial economics (for a recent review, see Bochkay et al 2023). We use the pretraining procedure proposed by Google and a large corpus of financial texts to develop FinBERT, an LLM specific to the finance domain.…”

“…Requiring a smaller training sample likely reduces the number of data labeling cycles and the costs associated with all these tasks. See Bochkay et al (2023) for a detailed discussion on the implementation guideline of training data for supervised machine learning.…”

The effect of social skills on analyst performance

“…1 Concurrent validity is the extent to which an operationalized construct correlates with theoretically related measures. Predictive validity is related to the ability of operationalized construct to predict an outcome that will result from the underlying theoretical construct (Bochkay et al, 2023).…”