Natural Language Processing is a rapidly evolving field with significant applications in various domains, especially in Finance. This talk will outline different approaches for analysing sentiments within various self-disclosures published on a company’s website. We will discuss the design and architecture of various techniques explored while developing the solution, including but not limited to dependency parsing, part-of-speech tagging, contextual understanding, and sentiment tagging.

The solution employs both classical machine learning techniques as well as Transformer-based models and we will cover how each component helped in the overall achievement of the desired performance. We will conclude with potential future developments, further research opportunities, and a Q&A session.