Many asset managers feel like they are drowning in information flow about their portfolios. Given the many sources of information moving asset prices - global news, market sentiment swings, and regulatory changes, among others - many turn to chatbot to distill news into digestible insights. And while general-purpose chatbot deliver useful insights, they can also deliver inaccurate or irrelevant content. Using NLP analytics to enhance AI accuracy, MarketPsych created a dedicated financial intelligence system that filters, synthesizes, and summarizes the information affecting individual portfolios.
A variety of formats for information that conveys “signal” (not “noise”) are available, as depicted below:
- Pre-market briefs - Comprehensive overnight summaries ready at market open
- Position-specific alerts - Real-time notifications on assets moving now, or subject to upcoming data releases, in a portfolio
- Industry intelligence - Emerging trends and themes in AI, biotech, renewable energy, and others
- Risk detection - Early warnings on regulatory changes, supply chain issues, geopolitical events
The financial intelligence system described in this paper identifies material and timely asset-related information, summarizes it, and presents it in a readable format that is customized to each user’s workflow. The system pipeline is depicted in the figure below.
Note the several stages and the feedback loop in the pipeline. The process relies upon accurate media data obtained from a repository of NLP-tagged global media sentences, which MarketPsych calls “Radar Database.”
An example of a summarized portfolio news brief – covering the Chinese Energy Storage industry - is below:
Example Portfolio News
ChatGPT vs NLP-Enhanced Summarization: The Key Differences
Summarizing recent news about the Chinese energy storage industry can also be performed by commercial chatbot such as ChatGPT. However, chatbot based on web search frequently fail to find and identify useful recent news into their summaries.
In the infographic below, we show differences between a standard chatbot-created news summarization service and one bolstered by NLP tagging. In Appendix 1 we provide a detailed table comparing the 2 approaches and in Appendix 2, we provide annotated screenshots of MarketPsych Summaries against ChatGPT in Exhibits A through D.