The goal of this project is to develop a predictive model using FinBERT to identify and forecast earnings call surprises. The project will build on the existing research on sentiment analysis of financial reports using NLP techniques, particularly on the use of FinBERT for analyzing earnings calls. The proposed model will take into account both the sentiment of the earnings call and the company's financial metrics leading up to the call to predict whether the call will result in a surprise or not. The model will be trained and tested on a dataset of earnings calls and their corresponding financial metrics to evaluate its performance. Ultimately, the model aims to provide valuable insights to investors, analysts, and other stakeholders in predicting the likelihood of earnings call surprises and informing investment decisions.