# Natural Language Commanding via Program Synthesis

**Authors:** Apurva Gandhi, Thong Q. Nguyen, Huitian Jiao, Robert Steen, Ameya Bhatawdekar  
**arXiv:** [2306.03460](https://arxiv.org/abs/2306.03460)  
**Date:** 2023-06-06  
**Summary page:** https://www.summarizepaper.com/en/paper/2306.03460v1/

## Key points

- Authors: Apurva Gandhi, Thong Q. Nguyen, Huitian Jiao, Robert Steen, Ameya Bhatawdekar
- Introduction of Semantic Interpreter: AI system enhancing productivity software like Microsoft Office
- Approach: Analysis-Retrieval prompt construction with large language models (LLMs) for program synthesis
- Process: Translating user natural language utterances into ODSL programs for application APIs execution
- Focus: Exploring Semantic Interpreter capabilities in Microsoft PowerPoint
- Goal: Streamlining and enhancing user interactions with productivity software through intuitive natural language commands

## Summary

In their paper titled "Natural Language Commanding via Program Synthesis," authors Apurva Gandhi, Thong Q. Nguyen, Huitian Jiao, Robert Steen, and Ameya Bhatawdekar introduce the Semantic Interpreter - an AI system designed to enhance productivity software like Microsoft Office by leveraging large language models (LLMs) to interpret user intent across various application features. The Semantic Interpreter utilizes an Analysis-Retrieval prompt construction approach in conjunction with LLMs for program synthesis. This process involves translating user utterances expressed in natural language into ODSL programs that can be transpiled into application APIs and subsequently executed. The focus of their research primarily centers around exploring the capabilities of Semantic Interpreter within Microsoft PowerPoint. By combining the power of LLMs with a domain-specific language like ODSL, the authors aim to streamline and enhance user interactions with productivity software by enabling seamless execution of complex tasks through intuitive natural language commands. Through their innovative approach, they pave the way for more efficient and user-friendly AI systems in the realm of productivity software.