Source: Waters Technology
This deep dive into AI from Waters Technology features Elena Treshcheva's (Researcher, Exactpro Systems) comments on the possibilities of applying recent NLP research to financial technology.
Natural language processing (NLP) is a well-established research direction applied to various knowledge domains, with the financial industry not being an exception.
In the financial services, natural language processing is leveraged to extract the information about the events happening globally and their perception by the people. This multidimensional information containing facts, opinions, sentiments, ranking, poll data, etc., is then projected onto the capital markets: the data provides analytical insights and the ability to predict stock market movements and identify potential risks. The most common areas for NLP application in finance are market surveillance and fraud detection, portfolio and risk management, conversational assistants in retail and investment banking.
One of the challenges of NLP being applied to solve complex tasks in finance is associated with domain-specific textual data availability. When it comes to training a model behind any AI application, it is not always easy to obtain sufficient business case specific datasets. Such approaches as BERT, developed by Google and available in open source, help to overcome this challenge.
BERT is a language model which is pre-trained on plain, unlabeled “general-purpose” textual data. The benefit of this model is that it can serve as a basis for building other, more task-specific ML models. In the case of AI-related financial tasks, this means a lot: BERT helps to perform the most difficult part of the task using non-specific text data and significantly contributes to a learning algorithm, which can subsequently be fine-tuned using domain-specific datasets (e.g. financial data) of a much smaller size.
Such developments in NLP like BERT are undoubtedly beneficial for a wider tech community. As a software testing specialist firm, Exactpro sees the application of this and other recently emerged approaches to quality assurance and their further enhancement as a strategically important activity. We already use various NLP methods in our work with complex mission-critical financial platforms. One direction of this work covers using AI in testing activities and includes, among others, log analysis for system monitoring, process mining for test coverage improvement, and test execution failures analysis for root cause predictions. Another direction is quality assurance of such AI-based systems as conversational assistants and machine-readable news. These systems are very often built using machine learning models, and we are adjusting our test approach to their verification and validation.
Some of these thoughts are featured in Waters Technology’s AI deep dive by Joanna Wright giving an overview of the recent NLP research applied to financial technology.