How Computers Use Memory Networks to Answer Questions
A method for AI to search through large amounts of stored information by repeatedly 'hopping' through memory to find the most relevant facts for answering a question.
Original patent title: “End-to-end memory networks”
A method for AI to search through large amounts of stored information by repeatedly 'hopping' through memory to find the most relevant facts for answering a question. Granted to Facebook Inc in 2020 with 16 claims and 1 forward citation.
Key facts
Coverage
What does this patent actually cover?
This patent describes a way for an AI model to process information by treating a database of knowledge like a long-term memory. When the system receives a question, it converts both the question and the stored facts into mathematical vectors. It then performs a 'hop' operation, where it calculates which facts are most relevant to the question using a probability score. By repeating this process across multiple hops, the system refines its focus, effectively 'reading' through the memory to build a better answer. For example, if you provided a story about a character moving between rooms, the system would use these hops to track the character's location before answering 'Where is the character now?'
The gap
What does this patent NOT cover?
- Does not cover traditional database lookups that rely on exact keyword matching.
- Does not cover systems that use hard-coded rules or logic trees to find answers.
- Does not cover non-vectorized storage methods for knowledge entries.
- Does not cover architectures that do not utilize a multi-hop iterative refinement process.
These exclusions are unique to PatentBrief — derived from the actual claim language, not patent-office boilerplate.
What made this novel
The system uses a 'continuous weighting function' (like softmax) during the hop process, which allows the model to be trained end-to-end using back-propagation. This means the AI learns how to search its own memory without needing human-labeled instructions on which specific facts are important.
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
Question-answering chatbots
Automated document summarization tools
Context-aware AI assistants
Why it matters
The bigger picture
This technology was a significant step in the evolution of Natural Language Processing (NLP) at Meta (formerly Facebook). It helped move AI beyond simple pattern matching toward models that can reason over long-term context, which is a foundational requirement for modern large language models.
Filed
March 28, 2017
Granted
May 26, 2020
Market context
Who's building on this
Companies in this space
Meta (Facebook) remains a primary developer of this technology, having integrated these concepts into their AI research division (FAIR). Many other major research labs and companies working on Retrieval-Augmented Generation (RAG) are building upon the fundamental idea of using memory-like structures to ground AI responses.
Market impact
This patent represents a shift toward more interpretable and grounded AI. By enabling models to look up external information rather than relying solely on internal parameters, it helped pave the way for the RAG architectures that are now standard in enterprise AI applications.
Claim 1 — Plain English
What this patent covers
This patent describes a way for an AI model to process information by treating a database of knowledge like a long-term memory. When the system receives a question, it converts both the question and the stored facts into mathematical vectors. It then performs a 'hop' operation, where it calculates which facts are most relevant to the question using a probability score. By repeating this process across multiple hops, the system refines its focus, effectively 'reading' through the memory to build a better answer. For example, if you provided a story about a character moving between rooms, the system would use these hops to track the character's location before answering 'Where is the character now?'
The clever bit
The system uses a 'continuous weighting function' (like softmax) during the hop process, which allows the model to be trained end-to-end using back-propagation. This means the AI learns how to search its own memory without needing human-labeled instructions on which specific facts are important.
What it does not cover
- Does not cover traditional database lookups that rely on exact keyword matching.
- Does not cover systems that use hard-coded rules or logic trees to find answers.
- Does not cover non-vectorized storage methods for knowledge entries.
- Does not cover architectures that do not utilize a multi-hop iterative refinement process.
Patent timeline
Application submitted to the patent office
Application published, typically 18 months after filing
Patent officially issued
PatentBrief Score
Impact Score
Moderate
Citation count
6/40
Early citations
Claim breadth
11/20
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
20/20
Major company or institution
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
$52K – $166K
Midpoint $104K · 10.8 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
16 claims as filed with the patent office.
Concepts involved
Citations
Patent lineage
Cite this patent
Sukhbaatar, S., Fergus, R. D., Weston, J. E., & Szlam, A. D. (2020). How Computers Use Memory Networks to Answer Questions (U.S. Patent No. 10,664,744). U.S. Patent and Trademark Office. https://patentbrief.org/patent/us/10664744/watson-question-answering-system-deepqa
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="US10664744"></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 AI & Machine Learning
US 10452978 · 2019 · Google LLC
How AI Models Understand Language Using 'Attention'
US 6523026 · 2003 · Huntsman International LLC
How Computers Find Hidden Connections Between Different Fields of Knowledge
US 11615208 · 2023 · Capital One Services LLC
How Cloud Systems Automatically Create and Train AI Data Models
US 10402750 · 2019 · Facebook Inc
How Facebook Uses Deep Learning to Predict What You Might Like
New to patents?
Common Questions
Frequently Asked Questions
What does How Computers Use Memory Networks to Answer Questions cover?
A method for AI to search through large amounts of stored information by repeatedly 'hopping' through memory to find the most relevant facts for answering a question.
Who owns patent US 10664744?
Facebook Inc owns this patent, granted in 2020.
When does this patent expire?
This patent is expected to expire on May 26, 2040, when the invention enters the public domain.
What is patent US 10664744 cited by?
This patent has been cited by 1 later patents that build on its ideas.
What problem does this patent solve?
This technology was a significant step in the evolution of Natural Language Processing (NLP) at Meta (formerly Facebook). It helped move AI beyond simple pattern matching toward models that can reason over long-term context, which is a foundational requirement for modern large language models.
What does this patent NOT cover?
Does not cover traditional database lookups that rely on exact keyword matching.
Same assignee
More from Facebook Inc
Patent monitoring