Natural Language Processing in Accounting, Auditing and Finance: A Synthesis of the Literature with a Roadmap for Future Research

Summary Behavioural science states that emotions, principles and the manner of thinking can affect the behaviour of individuals and even investors in their decision making on financial markets. In this paper, we have tried to measure the investor sentiment by three means of big data. The first is based on a search query of a list of words related to Islamic context. The second is inferred from the engagement degree on social media. The last measure of sentiment is built, based on the Twitter API classified into positive and negative directions by a machine learning algorithm based on the naive Bayes method. Then, we investigate whether these sensations and emotions have an impact on the market sentiment and the price fluctuations by means of a vector autoregression model and Granger causality analysis. In the final step, we apply the agent‐based simulation by means of the sequential Monte Carlo method with the control of our Twitter measure on Islamic index returns. We show, then, that the three social media sentiment measures present a remarkable impact on the contemporaneous and lagged returns of the different Islamic assets studied. We also give an estimation of the parameters of the latent variables relative to the agent model studied.

Summary We investigate the impact of social media data in predicting the Tehran Stock Exchange variables for the first time. We consider the closing price and daily return of three different stocks for this investigation. We collected our social media data from http://Sahamyab.com/stocktwits for about 3 months. To extract information from online comments, we propose a hybrid sentiment analysis approach that combines lexicon‐based and learning‐based methods. Since lexicons that are available for the Persian language are not practical for sentiment analysis in the stock market domain, we built a particular sentiment lexicon for this domain. After designing and calculating daily sentiment indices using the sentiment of the comments, we examine their impact on the baseline models that only use historical market data and propose new predictor models using multi‐regression analysis. In addition to the sentiments, we also examine the comments volume and the users' reliabilities. We conclude that the predictability of various stocks in the Tehran Stock Exchange is different depending on their attributes. Moreover, we indicate that only comments volume could be useful for predicting the closing price, and both the volume and the sentiment of the comments could be useful for predicting the daily return. We demonstrate that users' trust coefficients have different behaviours toward the three stocks.

Summary We implement an efficient methodology for extracting themes from Securities Exchange Commission 13D filings using aspects of human‐assisted active learning and long short‐term memory (LSTM) neural networks. Sentences from the ‘Purpose of Transaction’ section of each filing are extracted and a randomly chosen subset is labelled based on six filing themes that the existing literature on shareholder activism has shown to have an impact on stock returns. We find that an LSTM neural network that accepts sentences as input performs significantly better, with precision of 77%, than an alternately specified neural network that uses the common bag of words approach. This indicates that both sentence structure and vocabulary are important in classifying SEC 13D filings. Our study has important implications, as it addresses the recent cautions raised in the literature that analysis of finance and accounting‐related text sources should move beyond bag‐of‐words approaches to alternatives that incorporate the analysis of word sense and meaning reflecting context.