Houyu Zhang

Title: Houyu Zhang: Inventor of Language-Model Systems for Entity Representation.

Introduction

Houyu Zhang is an inventor based in Redmond, Washington, whose patent work explores how language models can combine different kinds of data to represent entities. He is associated with Amazon Technologies, Inc. His listed portfolio includes 1 patent.

Latest Patents

Zhang’s patent, “Systems for Determining Representations of Entities Based on Graph Data Using a Language Model,” describes a method for generating an overall representation of an entity with a large language model. The system draws on graph data about the entity and its relationships, text associated with it, and image data. It creates separate representations from these sources and uses them as input to a language model trained to produce a unified entity representation. Other models can then use that representation to identify entities with similar or different characteristics.

Career Highlights

The patent reflects work at the intersection of graph data, multimodal information, and language-model applications. It presents a way to combine structured relationships with textual and visual information for downstream analysis.

Collaborations

The patent information lists Seongjun Yun and Alejandro Ricardo Mottini d’Oliveira among Zhang’s coworkers.

Conclusion

Houyu Zhang’s patent highlights an approach to building richer entity representations by combining graph, text, and image data through a language model.