Paul Charles Tobola is an experienced financial professional and a regional vice president with OpenText. Paul Charles Tobola manages engagements for several global banks and works toward expanding these relationships to increase customer value. With over three decades of experience, Paul Tobola has established himself as an industry expert in finance.

Financial news provider Bloomberg has published a research paper that describes the creation of BloombergGPT, a language model specifically trained on financial data to support various natural language processing tasks in the financial sector. The model will enhance and streamline current financial natural language processing (NLP) tasks and enable Terminal clients to leverage the massive amounts of data in the market. Some of the tasks BloombergGPT can perform include news classification, sentiment analysis, named entity recognition, and question answering.

Engineers built BloombergGPT using a 363 billion token dataset of English financial documents, augmented with a 345 billion token public dataset from sources such as Wikipedia and YouTube. The resulting training corpus consists of over 700 billion tokens. Tokens are components of text that hold specific meanings in language processing. According to the head of Bloomberg's ML product and research team, Gideon Mann, BloombergGPT surpassed comparable open models in financial tasks by a significant margin and performed equally well or better on general natural language processing benchmarks at the time of the report.