From Code to Counsel: Deep Learning and Data Engineering Synergy for Intelligent Tax Strategy Generation

Abstract

Alas, despite the promise of "Artificial Intelligence" and the existing capabilities in raw data and neural network horsepower, the near-term reality for an open data and counsel-desiring government or corporation is that an explicit interaction of human consultants, data technicians, and tax lawyers is still a required part of such a "knowledge acquisition" process. This paper is directed toward mitigating those risks by describing a neural network approach to tax strategy formation, based on an aggregated corpus of knowledge providers and a gated generation network operating on labeled data constructed from that corpus. The combination models from legal and tax analysis branch this industry supplier domain document corpus and a novel stateful generation deep learning system to synthesize open domain services templates and semantics representing insights, patterns, strategies, and best practices for providing clients with analysis, insights, and counsel for different vertical and tax activities. With existing data and discerning data labeling, in collaboration with domain tax lawyers and data science, the "Artificial Intelligence" framework makes automatic tax strategy expressions possible.