Teaching Computers to Understand Document Similarity Using AI
This patent describes a way to train a computer program (a neural network) to understand how similar documents are to each other, by showing it examples and teaching it to group similar ones together and separate dissimilar ones.
Original patent title: “Content embedding using deep metric learning algorithms”
This patent describes a way to train a computer program (a neural network) to understand how similar documents are to each other, by showing it examples and teaching it to group similar ones together and separate dissimilar ones. Granted to Cognizant Technology Solutions US Corp in 2021 with 22 claims and 53 forward citations.
Key facts
Coverage
What does this patent actually cover?
This patent explains how to train a computer program, specifically a neural network, to create a 'space' where documents can be placed based on their meaning. Imagine you have a target document (like an article about dogs). You also give the program a 'favored' document (another article about dogs) and several 'unfavored' documents (articles about cats, cars, or anything else). The program learns by trying to make the 'dog' documents closer together in its 'space' and further away from the 'non-dog' documents. It does this by adjusting its internal settings, called parameters, to minimize a 'loss' function. This loss function measures how well it's separating the favored document from the unfavored ones relative to the target document. For instance, a training set might include an article about 'Golden Retrievers' (target), another about 'Labradors' (favored), and articles about 'Siamese Cats' and 'Electric Cars' (unfavored). The system adjusts itself so that the 'Golden Retriever' and 'Labrador' articles are 'close' in its internal representation, while the 'Siamese Cat' and 'Electric Car' articles are 'far' from the 'Golden Retriever' article.
The gap
What does this patent NOT cover?
- Does not cover methods that do not use a neural network for training.
- Does not cover training methods that do not involve a target document, a favored document, and at least two unfavored documents.
- Does not cover systems that do not calculate a 'loss' based on the distance between document representations.
- Does not cover methods where the computer program is not 'trained' using adjustable parameters.
- Does not cover creating an embedding space without using document vectors as input.
These exclusions are unique to PatentBrief — derived from the actual claim language, not patent-office boilerplate.
What made this novel
The core idea is teaching the AI not just to recognize what a document is about, but to learn the *relative similarity* between documents. By explicitly training it to bring 'good' matches closer and push 'bad' matches further away from a reference, it learns a nuanced understanding of meaning that's more effective than simply classifying documents.
Schematic visualization of the patent's claim structure. Hand-drawn diagrams in progress for each landmark patent.
Where you've seen this
Real-world examples
Search engine result ranking
Product recommendation systems
Content similarity detection
Plagiarism detection tools
Customer feedback analysis
Why it matters
The bigger picture
This technology is foundational for many modern AI applications that deal with understanding and organizing large amounts of text or other data. It enables search engines, recommendation systems, and content moderation tools to better grasp the meaning and relationships between different pieces of information.
Filed
June 9, 2017
Granted
February 2, 2021
Market context
Who's building on this
Companies in this space
Cognizant Technology Solutions, the assigneeassigneeThe entity that owns the patent — usually the inventor's employer or a company.Read more →, is a major IT services company that likely uses and builds upon such AI techniques for its clients. Many other tech companies, including major cloud providers like Google, Amazon, and Microsoft, develop and deploy similar deep metric learning algorithms for their search, recommendation, and AI services.
Market impact
This patent's approach to learning document embeddings has become a standard technique in the field of Natural Language Processing (NLP). It underpins the effectiveness of many modern search and recommendation engines, allowing them to provide more relevant results and suggestions by understanding semantic similarity rather than just keyword matching.
Claim 1 — Plain English
What this patent covers
This patent explains how to train a computer program, specifically a neural network, to create a 'space' where documents can be placed based on their meaning. Imagine you have a target document (like an article about dogs). You also give the program a 'favored' document (another article about dogs) and several 'unfavored' documents (articles about cats, cars, or anything else). The program learns by trying to make the 'dog' documents closer together in its 'space' and further away from the 'non-dog' documents. It does this by adjusting its internal settings, called parameters, to minimize a 'loss' function. This loss function measures how well it's separating the favored document from the unfavored ones relative to the target document. For instance, a training set might include an article about 'Golden Retrievers' (target), another about 'Labradors' (favored), and articles about 'Siamese Cats' and 'Electric Cars' (unfavored). The system adjusts itself so that the 'Golden Retriever' and 'Labrador' articles are 'close' in its internal representation, while the 'Siamese Cat' and 'Electric Car' articles are 'far' from the 'Golden Retriever' article.
The clever bit
The core idea is teaching the AI not just to recognize what a document is about, but to learn the *relative similarity* between documents. By explicitly training it to bring 'good' matches closer and push 'bad' matches further away from a reference, it learns a nuanced understanding of meaning that's more effective than simply classifying documents.
What it does not cover
- Does not cover methods that do not use a neural network for training.
- Does not cover training methods that do not involve a target document, a favored document, and at least two unfavored documents.
- Does not cover systems that do not calculate a 'loss' based on the distance between document representations.
- Does not cover methods where the computer program is not 'trained' using adjustable parameters.
- Does not cover creating an embedding space without using document vectors as input.
Patent timeline
Application submitted to the patent office
Application published, typically 18 months after filing
Patent officially issued
PatentBrief Score
Impact Score
Strong
Citation count
35/40
Highly cited
Claim breadth
15/20
Broad claimsclaimsThe numbered statements at the end of a patent that legally define what the inventor owns.Read more →
Recency
10/20
Granted 5–10 years ago
Assignee scale
0/20
Independent or smaller assigneeassigneeThe entity that owns the patent — usually the inventor's employer or a company.Read more →
PatentBrief Impact Score — based on citation count, claim breadth, recency, and assignee scale. Not a legal assessment.
Heuristic Value Estimate
What this patent might be worth
$468K – $1.5M
Midpoint $936K · 11.0 yr remaining · industry ×1.6
Heuristic only — blends forward/backward citation counts, claim scope, time remaining, litigation history, and CPC-derived industry baseline. Real valuations need a professional appraisal.
The original legal language
Original claims
22 claims as filed with the patent office.
Concepts involved
Citations
Patent lineage
Cite this patent
Legrand, D. G. M., Duffy, N., TSATSIN, P., & Long, P. M. (2021). Teaching Computers to Understand Document Similarity Using AI (U.S. Patent No. 10,909,459). U.S. Patent and Trademark Office. https://patentbrief.org/patent/us/10909459/federated-learning
Auto-generated from the patent record. Double-check author order and the issue date against the official USPTO document before submitting.
Embed
Add this patent to your site
Drop this plain-English patent card into any blog post or article — free, no signup. It always links back to the full breakdown here.
<div data-patentlens-widget data-patent-number="US10909459"></div> <script src="https://patentbrief.org/embed.js" async></script>
Stay in the loop
Get a weekly digest of new patents.
One email per week. No spam. Unsubscribe anytime.
Keep exploring
Related patents you should know
US 4683195 · 1987
How to Make Billions of Copies of a DNA Segment
This patent describes the Polymerase Chain Reaction (PCR), a method to rapidly create many copies of a specific piece of DNA or RNA, enabling its detection and analysis.
Cetus Corp
US 8697359 · 2014
How to Edit Genes in Human Cells Using an Engineered CRISPR System
This patent describes an engineered CRISPR-Cas9 system for precisely cutting DNA in eukaryotic cells to change how genes work, opening the door for gene editing in complex organisms.
Massachusetts Institute of Technology
US 7657849 · 2010
How the iPhone's Slide-to-Unlock Gesture Works
Apple's 2010 patent describes unlocking a device by dragging a specific graphical image across the touchscreen along a predefined path, a gesture that became iconic with the original iPhone.
Apple Inc
US 4733665 · 1988
How Doctors Implant a Permanent Stent Using a Balloon
This patent describes the method for placing a permanent, expandable wire mesh tube inside a blood vessel or other body tube using a balloon-tipped catheter to widen it and keep it open.
Expandable Grafts Partnership
US 4965188 · 1990
How to Make Many Copies of a DNA Piece with Heat
This patent describes the Polymerase Chain Reaction (PCR) method, a technique to make millions of copies of a specific DNA segment using a heat-resistant enzyme and repeated temperature changes.
Cetus Corp
US 4235871 · 1980
How to Encapsulate Active Materials in Lipid Bubbles Efficiently
This patent describes a method for trapping biologically active substances inside tiny, multi-layered fat bubbles called liposomes, using a specific water-in-oil emulsion and gel-forming process to improve how much material gets captured.
Individual
More to explore
More in Software & Internet
US 4405829 · 1983 · Massachusetts Institute of Technology
How RSA Public-Key Encryption Keeps Digital Messages Secret
US 6285999 · 2001 · Leland Stanford Junior University
How Websites Get Ranked by Importance
US 5960411 · 1999 · Amazon com Inc
How Amazon's One-Click Ordering Works for Online Purchases
US 7669123 · 2010 · Facebook Inc
Displaying Friends' Activities in a Social Network Feed
New to patents?
Common Questions
Frequently Asked Questions
What does Teaching Computers to Understand Document Similarity Using AI cover?
This patent describes a way to train a computer program (a neural network) to understand how similar documents are to each other, by showing it examples and teaching it to group similar ones together and separate dissimilar ones.
Who owns patent US 10909459?
Cognizant Technology Solutions US Corp owns this patent, granted in 2021.
When does this patent expire?
This patent is expected to expire on February 2, 2041, when the invention enters the public domain.
What is patent US 10909459 cited by?
This patent has been cited by 53 later patents that build on its ideas.
What problem does this patent solve?
This technology is foundational for many modern AI applications that deal with understanding and organizing large amounts of text or other data. It enables search engines, recommendation systems, and content moderation tools to better grasp the meaning and relationships between different pieces of information.
What does this patent NOT cover?
Does not cover methods that do not use a neural network for training.
Patent monitoring