# US 2024/0202440 A1 · Key points extraction for uniform resource locators

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

- Number: US 2024/0202440 A1
- Document ID: US20240202440A1
- Status: Published application
- Kind: A1
- Application number: 18/081,410
- Filed: 2022-12-14
- Published: 2024-06-20
- Earliest claimed: 2022-12-14
- Assignees: Microsoft Technology Licensing, LLC
- Applicants: Microsoft Technology Licensing, LLC
- Inventors: Mridu Baldevraj NARANG; Xuenan GUO; Shubham TRIPATHI; Guihong CAO; Zicheng HUANG; Xiaodong FAN
- Web page: https://patentgrep.com/patent/US20240202440A1

## CPC

- G06F 40/279: Recognition of textual entities
- G06F 40/12: Use of codes for handling textual entities
- G06N 3/08: Learning methods
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

The present disclosure relates to systems and methods for identifying webpage key points for a webpage. The systems and methods present the webpage key points in a side pane of a browser adjacent to the webpage.

## Claims

### Claim 1

A method, comprising:
receiving a uniform resource locator (URL) of a webpage presented on a browser;
determining that a selection threshold is satisfied for the webpage based on a webpage title of the webpage and content of the webpage;
identifying, using a machine learning model, webpage key points for the webpage, wherein the webpage key points are obtained from the webpage; and
causing the webpage key points to be presented adjacent to the webpage in a side pane of the browser.

### Claim 2 (depends on 1)

The method of claim 1, further comprising:
obtaining webpage sentences for the webpage;
mapping the webpage sentences to a distributed representation;
providing contextual interactions to the webpage sentences; and
predicting an importance of sentence spans of the webpage sentences.

### Claim 3 (depends on 2)

The method of claim 2, wherein identifying the webpage key points further comprises:
selecting sentence spans based on a ranking of the sentence spans; and
identifying the selected sentence spans as the webpage key points.

### Claim 4 (depends on 2)

The method of claim 2, wherein mapping the webpage sentences to the distributed representation further comprises:
generating sentence embeddings for the webpage sentences that are independent of a context of the webpage.

### Claim 5 (depends on 4)

The method of claim 4, wherein providing the contextual interactions to the webpage sentences further comprises:
fusing context information to the sentence embeddings; and
generating title-aware sentence embeddings for the webpage.

### Claim 6 (depends on 5)

The method of claim 5, wherein predicting the importance of the sentence spans of the webpage sentences further comprises:
generating proposed sentence spans using the title-aware sentence embeddings, wherein the proposed sentence spans provide a summary of content of the webpage.

### Claim 7 (depends on 6)

The method of claim 6, wherein ranking the sentence spans of the webpage sentences further comprises:
ranking the proposed sentence spans in descending order based on a confidence score of a summary provided by the proposed sentence spans; and
removing overlapping proposed sentence spans.

### Claim 8 (depends on 1)

The method of claim 1, further comprising:
selecting terms in the webpage key points; and
sending the terms and the webpage key points to be presented in the side pane of the browser, wherein a highlight is applied to the terms that visually distinguishes the terms in the webpage key points.

### Claim 9 (depends on 1)

The method of claim 1, further comprising:
comparing the webpage key points to a presentation threshold; and
sending, in response to the webpage key points being equal to the presentation threshold or exceeding the presentation threshold, the webpage key points for presentation in the side pane of the browser.

### Claim 10 (depends on 9)

The method of claim 9, wherein the presentation threshold identifies a confidence level in a summary provided by the webpage key points of the webpage.

### Claim 11 (depends on 9)

The method of claim 9, wherein the selection threshold controls whether the webpage key points are provided for the webpage and the presentation threshold controls whether the webpage key points are sent for presentation on the browser.

### Claim 12

A device, comprising:
a memory to store data and instructions; and
a processor operable to communicate with the memory, wherein the processor is operable to:
receive a uniform resource locator (URL) of a webpage presented on a browser;
determine that selection threshold is satisfied for the webpage based on a webpage title of the webpage and content of the webpage;
identify webpage key points for the webpage in response to the selection threshold being satisfied for the webpage, wherein the webpage key points are obtained from the webpage; and
cause the webpage key points to be presented in a side pane of a browser nearby the webpage.

### Claim 13 (depends on 12)

The device of claim 12, wherein the processor is further operable to use a plurality of machine learning models to identify the webpage key points.

### Claim 14 (depends on 12)

The device of claim 12, wherein the processor is further operable to identify webpage key points for the webpage by:
obtaining, from a datastore, sentence embeddings for the webpage and context information for the sentence embeddings;
generating title-aware sentence embeddings for the webpage using the sentence embeddings and the context information;
generating proposed sentence spans using the title-aware sentence embeddings, wherein the proposed sentence spans provide a summary of content of the webpage;
selecting sentence spans based on a ranking of the proposed sentence spans; and
identifying the selected sentence spans as the webpage key points.

### Claim 15 (depends on 12)

The device of claim 12, wherein the processor is further operable to:
select terms in the webpage key points; and
send the terms and the webpage key points for presentation in the side pane of the browser, wherein a highlight is applied to the terms that visually distinguishes the terms in the webpage key points.

### Claim 16 (depends on 12)

The device of claim 12, wherein the processor is further operable to:
compare the webpage key points to a presentation threshold that controls whether the webpage key points are sent for presentation on the browser; and
send, in response to the webpage key points being equal to the presentation threshold or exceeding the presentation threshold, the webpage key points for presentation in the side pane of the browser.

### Claim 17

A device, comprising:
a memory to store data and instructions; and
a processor operable to communicate with the memory, wherein the processor is operable to:
present, on a browser, a webpage and a side pane with an icon adjacent to the webpage;
send, in response to a selection of the icon, a uniform resource locator (URL) of the webpage to a summary model with a request for webpage key points;
receive, from the summary model based on the URL of the webpage, the webpage key points for the webpage, wherein the webpage key points are obtained from content of the webpage; and
present, in the side pane, the webpage key points in response to a selection of the icon.

### Claim 18 (depends on 17)

The device of claim 17, wherein the processor is further operable to:
provide a highlight of terms in the webpage key points that visually distinguishes the terms.

### Claim 19 (depends on 17)

The device of claim 17, wherein the summary model is a plurality of machine learning models that analyze webpage sentences of the webpage and a webpage title to determine the webpage key points.

### Claim 20 (depends on 17)

The device of claim 17, wherein the webpage key points are sentences or phrases obtained from different portions of the webpage.
