Author(s)

Yixuan Chen

Corresponding Author

Yixuan Chen

Abstract

Energy commodities, particularly oil and natural gas, exert systemic influence across industrial costs, international balances of trade, financial stability, and geopolitics. This study analyzes publicly available 2022–2024 sell-side research from two leading international investment banks and employs the large language model ERNIE Bot to extract topics and classify sentiments, thereby quantifying the relative importance of themes and mapping risk–opportunity profiles. Results indicate that market intelligence and price dynamics in oil and gas dominate the discourse; the Russia–Ukraine war is associated with strongly negative sentiment; policy and regulatory topics show mixed interpretations; and methodological themes such as trend factors, style integration, and the use of historical derivatives information are viewed positively. While sustainability and the green transition are increasingly present, they remain secondary to fossil-fuel fundamentals and event-driven factors. The findings demonstrate that large language models can convert unstructured financial text into structured insights that support strategy design, risk management, and forward-looking investment judgment in energy commodity markets.

Keywords

Energy Commodities; Oil and Natural Gas; Large Language Models; ERNIE Bot; Topic Modeling; Sentiment Analysis; Investment Strategy; Geopolitics