Microsoft Technology Licensing, LLC

Yarin Kuper

Royi Ronen

Erez Altus

Abedelkader Asi

Weizhu Chen

Roy Eisenstadt

Yi Mao

Alexander Tsvetkov

Abed El Kader Asi

Song Wang

Tomer Rosenthal

Rona Shaanan

Olga Goldenberg

Shimrit Rada Bemis

Example solutions for reducing the likelihood of hallucinations by language models, such as large language models (LLMs) are disclosed. By injecting a sufficient range and quantity of curated factual data into a prompt, the likelihood of a hallucination by an LLM may be reduced. This enables language models to be used in a wider range of settings, in which fabrication of facts is problematic, while reducing the need for a human to carefully check the generated text for accuracy. Examples include: generating a summary of a transcript using a summarization model; extracting topic-specific data from stored data using a scoring model; dynamically generating a language model prompt using the topic-specific data and the summary; and generating an output text using a language model and the language model prompt.

Dynamic construction of large language model prompts

The disclosure herein describes reducing training bias in outputs generated by a generative language model. A communication segment associated with a communication is obtained by at least one processor of a generative language model. An output value associated with the communication segment is generated by the generative language model. The output value is mapped to a set of training bias values associated with the generative language model and based on the mapping of the output value to a training bias value of the set of training bias values, an alternative output value is generated. The alternative output value is used in a generated segment output for the communication segment. The accuracy of segment outputs generated by the generative language model is improved through reducing or eliminating its training biases.

Reducing biases of generative language models

The disclosure herein describes determining topics of communication transcripts using trained summarization models. A first communication transcript associated with a first communication is obtained and divided into a first set of communication segments. A first set of topic descriptions is generated based on the first set of communication segments by analyzing each communication segment of the first set of communication segments with a generative language model. A summarization model is trained using the first set of communication segments and associated first set of topic descriptions as training data. The trained summarization model is then applied to a second communication transcript and, based on applying the trained summarization model to the second communication transcript, a second set of topic descriptions of the second communication transcript is generated. By training the summarization model based on output of the generative language model, it enables efficient, accurate generation of topic descriptions from communication transcripts.

Determining topic labels for communication transcripts based on a trained generative summarization model

Growing community of inventors

Tel Aviv, Israel

Yarin Kuper