# US 12,602,416 B2 · Hybrid artificial intelligence system for semi-automatic patent claims analysis

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 12,602,416 B2
- Document ID: US12602416B2
- Status: Granted
- Kind: B2
- Application number: 18/209,475
- Filed: 2023-06-13
- Published: 2023-10-12
- Granted: 2026-04-14
- Earliest claimed: 2019-08-20
- Assignees: Daystrom Information Systems, LLC
- Applicants: Daystrom Information Systems, LLC
- Inventors: Stephen G. Sullivan; Arnold E. Brown, II
- Web page: https://patentgrep.com/patent/US12602416B2

## CPC

- G06F 16/3344: {using natural language analysis}
- G06F 16/335: Filtering based on additional data, e.g. user or group profiles (filtering in web context G06F16/9535, G06F16/9536)
- G06F 3/0481: based on specific properties of the displayed interaction object or a metaphor-based environment, e.g. interaction with desktop elements like windows or icons, or assisted by a cursor's changing behaviour or appearance
- G06F 3/0482: Interaction with lists of selectable items, e.g. menus
- G06F 3/04847: Interaction techniques to control parameter settings, e.g. interaction with sliders or dials
- G06F 3/0486: Drag-and-drop
- G06F 16/338: Presentation of query results
- G06F 40/117: Tagging; Marking up (details of markup languages G06F40/143); Designating a block; Setting of attributes (style sheets, e.g. eXtensible Stylesheet Language Transformation [XSLT], G06F40/154)
- G06F 40/169: Annotation, e.g. comment data or footnotes
- G06F 40/279: Recognition of textual entities
- G06F 40/30: Semantic analysis
- G06N 3/0455: Auto-encoder networks; Encoder-decoder networks
- G06N 3/092: Reinforcement learning
- G06F 40/242: Dictionaries
- G06N 5/02: Knowledge representation; Symbolic representation
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

A user interface is displayed to enable a user to initiate a search based on a selected reference document. The UI displays options that enable the user to narrow the search, and the UI receives from the user a selection of portions of the reference document to search against and any annotations. A query is submitted to find a set of patents that are similar to the reference document, and a set of matching patents is received. Text of the reference document and the annotations are input to into a ML model to determine the meanings of words in the reference document. Text of the matching patents are input to the ML model to determine meanings of words in the matching patents. The reference document is searched to find matches between the claims from the matching patents and the reference document. A ranked set of the patents is then displayed.

## Claims

### Claim 1

A computer-implemented method for implementing patent claims analysis by a set of instructions, which when executed by a computer causes the computer to perform operations comprising:
displaying a user interface to enable a user to initiate a search based on a selected reference document;
displaying options in the user interface that enable the user to narrow the search, the user interface operable to receive from the user a selection of one or more portions of the reference document to search against, and any tagged annotations for the reference document;
submitting a query to find a set of patents that are semantical similar to the reference document;
receiving a set of matching patents from the query;
inputting text of the reference document and the tagged annotations into a machine learning (ML) model to determine meanings of a first set of words in the reference document;
inputting text of at least a portion of the matching patents to the ML model to determine meanings of a second set of words in the matching patents;
searching the one or more portions of the reference document with claims from the matching patents to find matches between the claims and the one or more portions of the reference document based on the meanings of the first set of words and the tagged annotations, and the meanings of the second set of words;
assigning confidence scores indicating a degree to which respective ones of the claims match the one or more portions of the reference document;
generating a ranked set of the patents based on the confidence scores; and
displaying identifiers of the ranked set of patents and the assigned confidence scores in the user interface.

### Claim 2 (depends on 1)

The computer-implemented method of claim 1, further comprising: training a set of different ML models corresponding to different technology domains, wherein each ML model is trained using a dataset of between 2,000 to 10,000 patents from the corresponding technology domain.

### Claim 3 (depends on 2)

The computer-implemented method of claim 2, further comprising: using one of the ML models to process the query that belongs to one of the different technology domains most closely related to a technology domain of the reference document.

### Claim 4 (depends on 3)

The computer-implemented method of claim 3, further comprising: prior to running the query, displaying in the user interface a list of the different technology domains; and receiving a user selection of a particular one of the different technology domains to use.

### Claim 5 (depends on 2)

The computer-implemented method of claim 2, further comprising: training each of the different ML models using a specific data set comprising words, phrases, and language associated with the corresponding technological domain.

### Claim 6 (depends on 5)

The computer-implemented method of claim 5, further comprising: using patents as the specific data set.

### Claim 7 (depends on 2)

The computer-implemented method of claim 2, further comprising: implementing the different technology domains as a plurality of global domains, and one or more user created custom domains, wherein the custom domains are trained using less data than the global domains and are created based upon information pulled from the reference document.

### Claim 8

A computer-implemented method for performing patent claims analysis by software component, which when executed by a computer, cause the computer to perform operations comprising:
displaying a user interface to enable a user to initiate a search based on a selected reference document;
displaying options in the user interface that enable the user to narrow a focus of the search, the user interface operable to receive from the user a selection of one or more portions of the reference document to search against, and any tagged annotations for the reference document;
submitting a query to find a set of patents that are semantical similar to the reference document;
receiving a set of matching patents from the query;
inputting text of the reference document and the tagged annotations into a trained machine learning (ML) model to transform reference document into reference document embeddings to determine meanings of a first set of words in the reference document;
inputting text of the matching patents to the trained ML model to transform the set of matching patents, including claim elements, into patent embeddings to determine the meanings of a second set of words in the matching patents;
for at least a portion of the patents, performing, by the trained ML model, an element-by-element search to find matches between the claim elements and the one or more portions of the reference document based on the meanings of the first set of words and the meanings of the second set of words;
ranking the set of patents based on assigned confidence scores indicating a degree to which the patent embeddings of the respective claim elements match the reference document embeddings of the
providing for viewing in the user interface a listing of the ranked set of patents, and for at least a portion of the set of patents, indications where matches occur between the claim elements and the one or more portions of the reference document.

### Claim 9 (depends on 8)

The computer-implemented method of claim 8, wherein the tagged annotations indicate i) meanings of any acronyms found in the reference document, ii) definitions of words, iii) text having increased weight applied during the search, or any combination thereof.

### Claim 10 (depends on 9)

The computer-implemented method of claim 9, further comprising: training a set of different ML models corresponding to different technology domains, wherein each ML model is trained using a dataset of between 2,000 to 10,000 patents from the corresponding technology domain.

### Claim 11 (depends on 10)

The computer-implemented method of claim 10, further comprising: using one of the ML models to process the query that belongs to one of the different technology domains most closely related to a technology domain of the reference document.

### Claim 12 (depends on 11)

The computer-implemented method of claim 11, further comprising: prior to running the query, displaying in the user interface a list of the different technology domains; and receiving a user selection of a particular one of the different technology domains to use.

### Claim 13 (depends on 10)

The computer-implemented method of claim 10, further comprising: training each of the different ML models using a specific data set comprising words, phrases, and language associated with the corresponding technological domain.

### Claim 14 (depends on 13)

The computer-implemented method of claim 13, further comprising: using patents as the specific data set.

### Claim 15 (depends on 10)

The computer-implemented method of claim 10, further comprising: implementing the different technology domains as a plurality of global domains, and one or more user-created custom domains, wherein the custom domains are trained using less data than the global domains and are created based upon information pulled from the reference document.

### Claim 16

An artificial intelligence (AI) system in communication with a client device over a network, the AI system to:
display a user interface to enable a user to initiate a search based on a selected reference document;
display options in the user interface that enable the user to narrow a focus of the search, the user interface operable to receive from the user a selection of one or more portions of the reference document to search against, and any tagged annotations for the reference document;
submit a query to find a set of patents that are semantical similar to the reference document;
receive a set of matching patents from the query;
input text of the reference document and the tagged annotations into a machine learning (ML) model to determine meanings of a first set of words in the reference document;
input text of at least a portion of the matching patents to the ML model to determine meanings of a second set of words in the matching patents;
search the one or more portions of the reference document with claims from the matching patents to find matches between the claims and the one or more portions of the reference document based on the meanings of the first set of words and the tagged annotations, and the meanings of the second set of words;
assign confidence scores indicating a degree to which respective ones of the claims match the one or more portions of the reference document;
generate a ranked set of the patents based on the confidence scores; and
display identifiers of the ranked set of patents and the assigned confidence A device may display a user interface to enable a user to initiate a search based on a selected reference document.