<scp>FinBERT</scp>: A Large Language Model for Extracting Information from Financial Text*

ABSTRACT We develop FinBERT, a state‐of‐the‐art large language model that adapts to the finance domain. show FinBERT incorporates knowledge and can better summarize contextual information in financial texts. Using sample of researcher‐labeled sentences from analyst reports, we document substantially outperforms Loughran McDonald dictionary other machine learning algorithms, including naïve Bayes, support vector machine, random forest, convolutional neural network, long short‐term memory, sentiment classification. Our results excels identifying positive or negative algorithms mislabel as neutral, likely because it uses text. find FinBERT's advantage over Google's original bidirectional encoder representations transformers model, is especially salient when training size small texts containing words not frequently used general also models discussions related environment, social, governance issues. Last, approaches underestimate textual informativeness earnings conference calls by at least 18% compared FinBERT. have implications for academic researchers, investment professionals, market regulators.