USPTO Patent Grant: Short Text Matching System

ChangeBridge: Patent Grants - AI & Computing (G06N)

Published March 24th, 2026

Detected March 25th, 2026

Summary

The USPTO has granted Intuit Inc. a patent for a system and method for short text matching using n-gram tokens and transformer models. The patent, effective March 24, 2026, aims to improve the accuracy of matching short text inputs like queries and target texts.

View original document View source feed page

What changed

The United States Patent and Trademark Office (USPTO) has granted patent US12585873B1 to Intuit Inc. for a "System and method for short text matching." This patent, effective March 24, 2026, details a novel approach utilizing pre-generated n-gram tokens and embeddings from a fine-tuned transformer model, processed by a one-layer transformer model for inference. The system is designed to enhance the accuracy of matching short text inputs, such as queries and target texts, by calculating similarity scores or classifications.

While this is a patent grant and not a regulatory rule imposing direct compliance obligations on businesses, it signifies a technological advancement in the field of AI and natural language processing. Companies operating in areas that involve text analysis, search, or classification may find this patented technology relevant for their product development or intellectual property strategy. No immediate compliance actions are required for regulated entities, but businesses developing or utilizing similar short text matching technologies should be aware of this granted patent and its potential implications for their intellectual property landscape.

Source document (simplified)

← USPTO Patent Grants

System and method for short text matching

Grant US12585873B1 Kind: B1 Mar 24, 2026

Assignee

Intuit Inc.

Inventors

Aleksandr Kim, Rineke van Noort, Yuting Lu, Ben Yi

Abstract

A short text matching system and method includes pre-generated dictionary of n-gram tokens having a selected length and corresponding embeddings produced by a fine-tuned transformer model and further includes a one-layer transformer model for inference. The dictionary is produced by fine-tuning a pretrained transformer model based on a domain specific short text training dataset. The length of the n-gram tokens is selected based on the dependency of the variance of embeddings on the n-gram length for embeddings produced by the fine-tuned transformer model. Domain specific input text, including query text and target text, are received and n-gram tokens of the selected length are produced. Embeddings corresponding to each of the n-gram tokens are determined from the dictionary along with corresponding positional embeddings. The n-gram embeddings and positional embeddings are provided to the one-layer transformer model, which produces a text matching result, such as similarity score or classification.