Daily Arxiv

This is a page that curates AI-related papers published worldwide.
All content here is summarized using Google Gemini and operated on a non-profit basis.
Copyright for each paper belongs to the authors and their institutions; please make sure to credit the source when sharing.

Agentic Retrieval of Topics and Insights from Earnings Calls

Created by
  • Haebom

Author

Anant Gupta, Rajarshi Bhowmik, Geoffrey Gunow

Outline

This paper presents a novel approach to analyze strategic focus by tracking topics that emerge in corporate quarterly earnings reports. We note that existing topic modeling techniques have difficulty in dynamically capturing emerging topics and their relationships as industries change, and propose a large-scale language model (LLM) agent-based approach. The LLM agent extracts topics from documents, structures them into a hierarchical ontology, and establishes relationships between new and existing topics within the ontology. We use the extracted topics to infer firm-level insights and emerging trends over time, and evaluate the approach by measuring ontology consistency, topic evolution accuracy, and the ability to suggest new financial trends.

Takeaways, Limitations

•
Takeaways:
◦
We present a novel way to effectively extract and analyze dynamically changing topics from corporate earnings reports using LLM Agent.
◦
By building a hierarchical ontology, you can clearly understand the relationships between topics and analyze the changes in a company's strategic focus over time.
◦
Extracted topics can be used to derive enterprise-level insights and new financial trends.
•
Limitations:
◦
Further validation is needed on the objectivity and generalizability of the metrics (e.g. ontology consistency, topic evolution accuracy, etc.) used to evaluate the performance of the proposed LLM agent.
◦
Since the LLM agent is highly dependent on training data, there is a possibility that the results may be unreliable due to data bias.
◦
It may not fully reflect the complexity of actual financial markets and expert judgment may be required to interpret the extracted topics.
👍
Made with Slashpage