Abstract
Business decision-makers rely on a causal map between strategy and performance to make good decisions. However, the causal insights about a business or industry are often fragmented and disconnected across various reporting and research documents. For instance, the International Federation of Accountants estimates the efforts of integrating insights from different reporting documentsfor decision making may have cost the financial industry alone $780 billion annually. Our proposed research seeks to address this problem by automating the detection and integration of key insights from textual data into a causal knowledge graph for data analytics and decision-making. We propose to develop a prototype of machine reading to detect, deconstruct, and integrate causal propositions from scholarly and reporting texts into a knowledge graph data, showing all the causal pathways among strategy-relevant variables and enterprise performance outcomes. Specifically, we will develop a prototype combining multiple machine-reading models, and train them on a sample of recent SEC documents of S&P financial companies. With an expert from the financial industry, we are developing a proposal for NSF PFI-RP grant due July 14, 2021, as well as pursuing industry sponsorship.
