Abstract
Presented herein are techniques for improving Conversational Artificial Intelligence (AI) systems through the incorporation of domain expert knowledge, leveraging the Retrieval-Augmented Generation (RAG) process. Techniques described herein provide an approach that begins with the establishment of semantic relationships within the questions, choices, and solutions components of a decision tree. Semantically-structured questions are sent to GPT-4 for keyword and semantic tagging, incorporating domain-specific information. After integrating these new tags, the enriched dataset is again processed by GPT-4 for natural language translation. This process can be utilized for RAG, model building, prompt engineering, and fine-tuning. The final output is utilized by applications that leverage GPT-based text embedding, significantly enhancing context-aware interactions and domain-specific information retrieval.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Teeter, James J. II, Ph.D., "SEMANTIC-FIRST DECISION TREE NATURALIZATION FOR ENHANCED CONVERSATIONAL ARTIFICIAL INTELLIGENCE", Technical Disclosure Commons, (July 25, 2024)
https://www.tdcommons.org/dpubs_series/7234