Will WalkerFollow


Virtual assistant applications and other artificial intelligence (AI) agents can use large language models (LLMs) to generate responses to spoken user queries. However, responses from LLMs can be long and include more information than the user’s specific need. As a result, it can be difficult for users to understand and remember the answers easily simply by listening. This disclosure describes techniques to enhance the conversational capabilities of voice based interactions with virtual assistant applications and other AI agents by enabling users to interrupt spoken responses with a new query. The context of the interruption, including contents of the spoken response from the AI agent and the timing of the interruption, and the user’s new query are provided to the LLM. The LLM can utilize the context to provide specific responses to the user’s query, thus providing a natural conversational experience. Incorporating the content and context of the interruption in subsequent responses can significantly improve the accuracy and quality of the responses.

Creative Commons License

Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.