Financial text analytics

Text consists of the majority of alternative data. Analyzing the versatile financial text coming from different sources, including corporate disclosures, annual reports, earning calls, financial news and social media, however, presents challenges. This thread of research designs NLP methods and investigates their adaptation to the various language domains in finance.

Representative findings are:

• In predictive analytics, text source and predictive horizon are two primary factors to consider.
• Parallel knowledge integration is the most effective approach to build a neuro-symbolic model for financial sentiment analysis.

Intelligent finance models

This stream of research is dual to financial text analytics: we live in a world where financial services and markets are digitalizing, providing an unprecedented amount of information to support decision-making. Traditional models in finance, developed decades ago, has overlooked such unstructured information. By leveraging on the methods developed to monitor public moods and discussions. This new information can be integrated to the financial forecasting and investing models to help improve them. An ongoing project in this stream is:

• Digitalization of the modern portfolio theory (MPT).

Knowledge discovery and engineering

Understanding financial texts requires dealing with a lot of numbers, time expressions, and contextualized arguments, many of which are not possible without external resources. Well-curated knowledge is therefore, always important. Today, most NLP studies benifit, directly or indirectly, from early efforts such as WordNet and DBpedia. In that sense, knowledge is both the purpose and the instrument (ct. Sukhomlinskii).

Showcases under this umbrella include NEEQ business taxonomy, FinSenticNet, and a lexicon domain-adaptation method.

Green IT and green finance

Certain economic activities and developments have negative externality: a typical example is environmental pollution, and climate change, when it becomes a global problem. The way how we use technology makes a huge difference. The first wave of IT, for instance office automation, had many good environmental impacts. But with the recent prevalence of AI, we should be cautious about low efficiency or even destructive computing. Sometimes, unregulated use can turn a powerful invention into an existential threat. Some key questions to ask are:

• How can text mining be used to discover and support greener and innovative ideas?