ABSTRACT
Investment research involves collecting diverse information from various sources, and piecing these together to form a mosaic; this mosaic is then used to obtain a view of the risks and opportunities in the market or within an asset class. The task is laborious, prone to bias and human error and naturally limited in its scope. In our implementation of the Global Economy and Markets Sentiment (GEMS) model, we have considered news sentiment analysis and have addressed these shortcomings. We have reached into previously untapped information sources and employed the tools of AI and machine learning, and power computing to manage scale while mitigating errors and biases. In this paper, we describe the model's core framework and implementation, and the dataset upon which it was built. We propose different use cases and explore several investment strategies; this work is based on insights obtained by us through our continuing and sustained research. The model was conceived and developed within the Managed Investments department at the Northwestern Mutual Life Insurance Company, specifically by the department's Quantitative Research and Strategy team. In its current application, the GEMS model helps the Emerging Markets investment team gain insights from large volumes of global and local news in over 65 languages. The principal objective is to reach better and informed investment decisions and to achieve these before the market moves.
