Natural language processing (NLP), or the pragmatic research perspective of computational linguistics, has become increasingly powerful due to data availability and various techniques developed in the past decade. This increasing capability makes it possible to capture sentiments more accurately and semantics in a more nuanced way. Naturally, many applications are starting to seek improvements by adopting cutting-edge NLP techniques. Financial forecasting is no exception. As a result, articles that leverage NLP techniques to predict financial markets are fast accumulating, gradually establishing the research field of natural language based financial forecasting (NLFF), or from the application perspective, stock market prediction. This review article clarifies the scope of NLFF research by ordering and structuring techniques and applications from related work. The survey also aims to increase the understanding of progress and hotspots in NLFF, and bring about discussions across many different disciplines.

Deep learning has unlocked new paths towards the emulation of the peculiarly-human capability of learning from examples. While this kind of bottom-up learning works well for tasks such as image classification or object detection, it is not as effective when it comes to natural language processing. Communication is much more than learning a sequence of letters and words: it requires a basic understanding of the world and social norms, cultural awareness, commonsense knowledge, etc.; all things that we mostly learn in a top-down manner. In this work, we integrate top-down and bottom-up learning via an ensemble of symbolic and subsymbolic AI tools, which we apply to the interesting problem of polarity detection from text. In particular, we integrate logical reasoning within deep learning architectures to build a new version of SenticNet, a commonsense knowledge base for sentiment analysis.

Financial Sentiment Analysis (FSA) is an important domain application of sentiment analysis that has gained increasing attention in the past decade. FSA research falls into two main streams. The first stream focuses on defining tasks and developing techniques for FSA, and its main objective is to improve the performances of various FSA tasks by advancing methods and using/curating human-annotated datasets. The second stream of research focuses on using financial sentiment, implicitly or explicitly, for downstream applications on financial markets, which has received more research efforts. The main objective is to discover appropriate market applications for existing techniques. More specifically, the application of FSA mainly includes hypothesis testing and predictive modeling in financial markets. This survey conducts a comprehensive review of FSA research in both the technique and application areas and proposes several frameworks to help understand the two areas’ interactive relationship. This article defines a clearer scope for FSA studies and conceptualizes the FSA-investor sentiment-market sentiment relationship. Major findings, challenges, and future research directions for both FSA techniques and applications have also been summarized and discussed.

Large language models (LLMs) have drastically changed the possible ways to design intelligent systems, shifting the focus from massive data acquisition and new model training to human alignment and strategic elicitation of the full potential of existing pre-trained models. This paradigm shift, however, is not fully realized in financial sentiment analysis (FSA) due to the discriminative nature of this task and a lack of prescriptive knowledge of how to leverage existing generative models in such a context. This study investigates the effectiveness of the new paradigm, that is, using LLMs without fine-tuning for FSA. Rooted in Minsky’s theory of mind and emotions, a design framework with heterogeneous LLM agents is proposed and applied to FSA. The framework instantiates specialized agents using prior guiding knowledge from both linguistics and finance. Then, a summative agent reasons on the aggregated agent discussions. Comprehensive evaluations using six FSA datasets show that the framework yields better accuracies compared to many alternative multi-LLM agent settings, especially when the discussion contents are substantial. This study contributes to the design foundations and paves new avenues for LLMs-based FSA and potentially other tasks. Implications for business and management have also been discussed.