# US 11,308,426 B2 · Sequence modeling for searches

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

- Number: US 11,308,426 B2
- Document ID: US11308426B2
- Status: Granted
- Kind: B2
- Application number: 16/234,377
- Filed: 2018-12-27
- Published: 2020-07-02
- Granted: 2022-04-19
- Earliest claimed: 2018-12-27
- Assignees: Microsoft Technology Licensing, LLC
- Applicants: Microsoft Technology Licensing, LLC
- Inventors: Meng Meng; Haifeng Zhao
- Web page: https://patentgrep.com/patent/US11308426B2

## CPC

- G06Q 10/063112: {Skill-based matching of a person or a group to a task}
- G06F 16/9535: Search customisation based on user profiles and personalisation
- G06N 5/04: Inference or reasoning models
- G06N 20/00: Machine learning
- G06Q 10/1053: Employment or hiring
- G06Q: INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

The disclosed embodiments provide a system for performing sequence modeling for searches. During operation, the system obtains a sequence of jobs associated with activity by a member of an online system. Next, the system applies a word embedding model of a set of job histories to attributes of individual jobs in the sequence of jobs to produce embeddings for the individual jobs. The system then generates a set of power means from the embeddings. Finally, the system outputs the set of power means as an encoded representation of the sequence of jobs, wherein the set of power means is used in generating job recommendations related to the member.

## Claims

### Claim 1

A computer-implemented method, comprising:
using a word embedding model, generating word embeddings for words used as job attributes, the word embedding model trained to generate the word embeddings with training data comprising job attributes obtained from member profiles of members of an online system, the job attributes comprising one or more of job titles, industries, company names, schools, and/or fields of study;
processing a first member profile of a first member of the online system to identify one or more words associated with each job in a sequence of jobs associated with the first member;
converting the one or more words associated with each job in the sequence of jobs to a pre-trained word embedding generated by the word embedding model;
generating a set of power means from the pre-trained word embeddings; and
outputting the set of power means as a fixed length encoded representation of the sequence of jobs.

### Claim 2 (depends on 1)

The method of claim 1, wherein generating the set of power means from the pre-trained word embeddings comprises:
calculating the set of power means from the pre-trained word embeddings based on a set of parameter values for a power mean function.

### Claim 3 (depends on 2)

The method of claim 2, wherein the set of parameter values comprises:
negative infinity;
positive infinity;
one; and
an odd number greater than one.

### Claim 4 (depends on 1)

The method of claim 1, further comprising:
training the word embedding model to generate the word embeddings for words used as job attributes based on groupings of job attributes that represent the sequence of jobs.

### Claim 5 (depends on 4)

The method of claim 4, wherein the job attributes in the grouping of job attributes comprise:
a previous title;
a current title; and
a company.

### Claim 6 (depends on 4)

The method of claim 4, wherein the job attributes in the grouping of job attributes comprise:
a school;
a field of study; and
an industry.

### Claim 7 (depends on 1)

The method of claim 1, further comprising:
providing as input to a machine learning model the fixed length encoded representation of the sequence of jobs and an additional embedding for a job associated with a job listing hosted by the online system;
receiving, as output from the machine learning model, a score representing a likelihood of a positive response to the job associated with the job listing by the first member; and
generating a recommendation of the job associated with the job listing to the first member based on the score.

### Claim 8 (depends on 1)

The method of claim 1, wherein the sequence of jobs comprises at least one of:
a job history of the first member;
a job search history of the first member; and
a job application history of the first member.

### Claim 9 (depends on 1)

The method of claim 1, wherein the job attributes comprise at least one of:
a title;
a company; and
an industry.

### Claim 10

A system, comprising:
one or more processors; and
memory storing instructions that, when executed by the one or more processors, cause the system to:
using a word embedding model, generate word embeddings for words used as job attributes, the word embedding model trained to generate the word embeddings with training data comprising job attributes obtained from member profiles of members of an online system, the job attributes comprising one or more of job titles, industries, company names, schools, and/or fields of study;
process a first member profile of a first member of the online system to identify one or more words associated with each job in a sequence of jobs associated with the first member;
convert the one or more words associated with each job in the sequence of jobs to a pre-trained word embedding generated by the word embedding model;
generate a set of power means from the pre-trained word embeddings; and
output the set of power means as a fixed length encoded representation of the sequence of jobs.

### Claim 11 (depends on 10)

The system of claim 10, wherein generating the set of power means from the pre-trained word embeddings comprises:
calculating the set of power means from the pre-trained word embeddings based on a set of parameter values for a power mean function.

### Claim 12 (depends on 11)

The system of claim 11, wherein the set of parameter values comprises:
negative infinity;
positive infinity;
one; and
an odd number greater than one.

### Claim 13 (depends on 10)

The system of claim 10, wherein the memory further stores instructions that, when executed by the one or more processors, cause the system to:
train the word embedding model to generate the word embeddings for words used as job attributes based on groupings of job attributes that represent the sequence of jobs.

### Claim 14 (depends on 10)

The system of claim 10, further comprising:
providing as input to a machine learning model the fixed length encoded representation of the sequence of jobs and an additional embedding of a job associated with a job listing hosted by the online system;
receiving, as output from the machine learning model, a score representing a likelihood of a positive response to the job associated with the job listing by the first member; and
generating a recommendation of the job associated with the job listing to the first member based on the score.

### Claim 15 (depends on 10)

The system of claim 10, wherein the sequence of jobs comprises at least one of:
a job history of the first member;
a job search history of the first member; and
a job application history of the first member.

### Claim 16

A non-transitory computer-readable storage medium storing instructions that when executed by a computer cause the computer to perform a method, the method comprising:
using a word embedding model, generating word embeddings for words used as job attributes, the word embedding model trained to generate the word embeddings with training data comprising job attributes obtained from member profiles of members of an online system, the job attributes comprising one or more of job titles, industries, company names, schools, and/or fields of study;
processing a first member profile of a first member of the online system to identify one or more words associated with each job in a sequence of jobs associated with the first member;
converting the one or more words associated with each job in the sequence of jobs to a pre-trained word embedding generated by the word embedding model;
generating a set of power means from the pre-trained word embeddings; and
outputting the set of power means as a fixed length encoded representation of the sequence of jobs.